text
stringlengths 29
3.31k
| label
sequencelengths 1
11
|
|---|---|
In this thesis, we present new schemes which leverage a constrained
clustering method to solve several computer vision tasks ranging from image
retrieval, image segmentation and co-segmentation, to person re-identification.
In the last decades clustering methods have played a vital role in computer
vision applications; herein, we focus on the extension, reformulation, and
integration of a well-known graph and game theoretic clustering method known as
Dominant Sets. Thus, we have demonstrated the validity of the proposed methods
with extensive experiments which are conducted on several benchmark datasets. | [
"cs.CV"
] |
While self-supervised monocular depth estimation in driving scenarios has
achieved comparable performance to supervised approaches, violations of the
static world assumption can still lead to erroneous depth predictions of
traffic participants, posing a potential safety issue. In this paper, we
present R4Dyn, a novel set of techniques to use cost-efficient radar data on
top of a self-supervised depth estimation framework. In particular, we show how
radar can be used during training as weak supervision signal, as well as an
extra input to enhance the estimation robustness at inference time. Since
automotive radars are readily available, this allows to collect training data
from a variety of existing vehicles. Moreover, by filtering and expanding the
signal to make it compatible with learning-based approaches, we address radar
inherent issues, such as noise and sparsity. With R4Dyn we are able to overcome
a major limitation of self-supervised depth estimation, i.e. the prediction of
traffic participants. We substantially improve the estimation on dynamic
objects, such as cars by 37% on the challenging nuScenes dataset, hence
demonstrating that radar is a valuable additional sensor for monocular depth
estimation in autonomous vehicles. Additionally, we plan on making the code
publicly available. | [
"cs.CV",
"cs.LG",
"cs.RO"
] |
We propose a novel algorithm, named Open-Edit, which is the first attempt on
open-domain image manipulation with open-vocabulary instructions. It is a
challenging task considering the large variation of image domains and the lack
of training supervision. Our approach takes advantage of the unified
visual-semantic embedding space pretrained on a general image-caption dataset,
and manipulates the embedded visual features by applying text-guided vector
arithmetic on the image feature maps. A structure-preserving image decoder then
generates the manipulated images from the manipulated feature maps. We further
propose an on-the-fly sample-specific optimization approach with
cycle-consistency constraints to regularize the manipulated images and force
them to preserve details of the source images. Our approach shows promising
results in manipulating open-vocabulary color, texture, and high-level
attributes for various scenarios of open-domain images. | [
"cs.CV"
] |
Recent success in deep reinforcement learning for continuous control has been
dominated by model-free approaches which, unlike model-based approaches, do not
suffer from representational limitations in making assumptions about the world
dynamics and model errors inevitable in complex domains. However, they require
a lot of experiences compared to model-based approaches that are typically more
sample-efficient. We propose to combine the benefits of the two approaches by
presenting an integrated approach called Curious Meta-Controller. Our approach
alternates adaptively between model-based and model-free control using a
curiosity feedback based on the learning progress of a neural model of the
dynamics in a learned latent space. We demonstrate that our approach can
significantly improve the sample efficiency and achieve near-optimal
performance on learning robotic reaching and grasping tasks from raw-pixel
input in both dense and sparse reward settings. | [
"cs.LG",
"cs.AI",
"cs.RO",
"stat.ML"
] |
We present FusedGAN, a deep network for conditional image synthesis with
controllable sampling of diverse images. Fidelity, diversity and controllable
sampling are the main quality measures of a good image generation model. Most
existing models are insufficient in all three aspects. The FusedGAN can perform
controllable sampling of diverse images with very high fidelity. We argue that
controllability can be achieved by disentangling the generation process into
various stages. In contrast to stacked GANs, where multiple stages of GANs are
trained separately with full supervision of labeled intermediate images, the
FusedGAN has a single stage pipeline with a built-in stacking of GANs. Unlike
existing methods, which requires full supervision with paired conditions and
images, the FusedGAN can effectively leverage more abundant images without
corresponding conditions in training, to produce more diverse samples with high
fidelity. We achieve this by fusing two generators: one for unconditional image
generation, and the other for conditional image generation, where the two
partly share a common latent space thereby disentangling the generation. We
demonstrate the efficacy of the FusedGAN in fine grained image generation tasks
such as text-to-image, and attribute-to-face generation. | [
"cs.CV"
] |
Generative Adversarial Networks (GAN) have demonstrated impressive results in
modeling the distribution of natural images, learning latent representations
that capture semantic variations in an unsupervised basis. Beyond the
generation of novel samples, it is of special interest to exploit the ability
of the GAN generator to model the natural image manifold and hence generate
credible changes when manipulating images. However, this line of work is
conditioned by the quality of the reconstruction. Until now, only inversion to
the latent space has been considered, we propose to exploit the representation
in intermediate layers of the generator, and we show that this leads to
increased capacity. In particular, we observe that the representation after the
first dense layer, present in all state-of-the-art GAN models, is expressive
enough to represent natural images with high visual fidelity. It is possible to
interpolate around these images obtaining a sequence of new plausible synthetic
images that cannot be generated from the latent space. Finally, as an example
of potential applications that arise from this inversion mechanism, we show
preliminary results in exploiting the learned representation in the attention
map of the generator to obtain an unsupervised segmentation of natural images. | [
"cs.CV",
"cs.LG",
"stat.ML"
] |
Pruning the parameters of deep neural networks has generated intense interest
due to potential savings in time, memory and energy both during training and at
test time. Recent works have identified, through an expensive sequence of
training and pruning cycles, the existence of winning lottery tickets or sparse
trainable subnetworks at initialization. This raises a foundational question:
can we identify highly sparse trainable subnetworks at initialization, without
ever training, or indeed without ever looking at the data? We provide an
affirmative answer to this question through theory driven algorithm design. We
first mathematically formulate and experimentally verify a conservation law
that explains why existing gradient-based pruning algorithms at initialization
suffer from layer-collapse, the premature pruning of an entire layer rendering
a network untrainable. This theory also elucidates how layer-collapse can be
entirely avoided, motivating a novel pruning algorithm Iterative Synaptic Flow
Pruning (SynFlow). This algorithm can be interpreted as preserving the total
flow of synaptic strengths through the network at initialization subject to a
sparsity constraint. Notably, this algorithm makes no reference to the training
data and consistently competes with or outperforms existing state-of-the-art
pruning algorithms at initialization over a range of models (VGG and ResNet),
datasets (CIFAR-10/100 and Tiny ImageNet), and sparsity constraints (up to
99.99 percent). Thus our data-agnostic pruning algorithm challenges the
existing paradigm that, at initialization, data must be used to quantify which
synapses are important. | [
"cs.LG",
"cond-mat.dis-nn",
"cs.CV",
"q-bio.NC",
"stat.ML"
] |
Traditional human activity recognition (HAR) based on time series adopts
sliding window analysis method. This method faces the multi-class window
problem which mistakenly labels different classes of sampling points within a
window as a class. In this paper, a HAR algorithm based on U-Net is proposed to
perform activity labeling and prediction at each sampling point. The activity
data of the triaxial accelerometer is mapped into an image with the single
pixel column and multi-channel which is input into the U-Net network for
training and recognition. Our proposal can complete the pixel-level gesture
recognition function. The method does not need manual feature extraction and
can effectively identify short-term behaviors in long-term activity sequences.
We collected the Sanitation dataset and tested the proposed scheme with four
open data sets. The experimental results show that compared with Support Vector
Machine (SVM), k-Nearest Neighbor (kNN), Decision Tree(DT), Quadratic
Discriminant Analysis (QDA), Convolutional Neural Network (CNN) and Fully
Convolutional Networks (FCN) methods, our proposal has the highest accuracy and
F1-socre in each dataset, and has stable performance and high robustness. At
the same time, after the U-Net has finished training, our proposal can achieve
fast enough recognition speed. | [
"cs.LG",
"cs.AI",
"stat.ML"
] |
Cross-spectral iris recognition is emerging as a promising biometric approach
to authenticating the identity of individuals. However, matching iris images
acquired at different spectral bands shows significant performance degradation
when compared to single-band near-infrared (NIR) matching due to the spectral
gap between iris images obtained in the NIR and visual-light (VIS) spectra.
Although researchers have recently focused on deep-learning-based approaches to
recover invariant representative features for more accurate recognition
performance, the existing methods cannot achieve the expected accuracy required
for commercial applications. Hence, in this paper, we propose a conditional
coupled generative adversarial network (CpGAN) architecture for cross-spectral
iris recognition by projecting the VIS and NIR iris images into a
low-dimensional embedding domain to explore the hidden relationship between
them. The conditional CpGAN framework consists of a pair of GAN-based networks,
one responsible for retrieving images in the visible domain and other
responsible for retrieving images in the NIR domain. Both networks try to map
the data into a common embedding subspace to ensure maximum pair-wise
similarity between the feature vectors from the two iris modalities of the same
subject. To prove the usefulness of our proposed approach, extensive
experimental results obtained on the PolyU dataset are compared to existing
state-of-the-art cross-spectral recognition methods. | [
"cs.CV"
] |
An autoencoder is a layered neural network whose structure can be viewed as
consisting of an encoder, which compresses an input vector of dimension $D$ to
a vector of low dimension $d$, and a decoder which transforms the
low-dimensional vector back to the original input vector (or one that is very
similar). In this paper we explore the compressive power of autoencoders that
are Boolean threshold networks by studying the numbers of nodes and layers that
are required to ensure that the numbers of nodes and layers that are required
to ensure that each vector in a given set of distinct input binary vectors is
transformed back to its original. We show that for any set of $n$ distinct
vectors there exists a seven-layer autoencoder with the smallest possible
middle layer, (i.e., its size is logarithmic in $n$), but that there is a set
of $n$ vectors for which there is no three-layer autoencoder with a middle
layer of the same size. In addition we present a kind of trade-off: if a
considerably larger middle layer is permissible then a five-layer autoencoder
does exist. We also study encoding by itself. The results we obtain suggest
that it is the decoding that constitutes the bottleneck of autoencoding. For
example, there always is a three-layer Boolean threshold encoder that
compresses $n$ vectors into a dimension that is reduced to twice the logarithm
of $n$. | [
"cs.LG",
"stat.ML"
] |
We present a new approach to 3D object representation where a neural network
encodes the geometry of an object directly into the weights and biases of a
second 'mapping' network. This mapping network can be used to reconstruct an
object by applying its encoded transformation to points randomly sampled from a
simple geometric space, such as the unit sphere. We study the effectiveness of
our method through various experiments on subsets of the ShapeNet dataset. We
find that the proposed approach can reconstruct encoded objects with accuracy
equal to or exceeding state-of-the-art methods with orders of magnitude fewer
parameters. Our smallest mapping network has only about 7000 parameters and
shows reconstruction quality on par with state-of-the-art object decoder
architectures with millions of parameters. Further experiments on feature
mixing through the composition of learned functions show that the encoding
captures a meaningful subspace of objects. | [
"cs.LG",
"cs.CV",
"cs.RO",
"stat.ML"
] |
We address the issue of creating consistent mesh texture maps captured from
scenes without color calibration. We find that the method for aggregation of
the multiple views is crucial for creating spatially consistent meshes without
the need to explicitly optimize for spatial consistency. We compute a color
prior from the cross-correlation of observable view faces and the faces per
view to identify an optimal per-face color. We then use this color in a
re-weighting ratio for the best-view texture, which is identified by prior mesh
texturing work, to create a spatial consistent texture map. Despite our method
not explicitly handling spatial consistency, our results show qualitatively
more consistent results than other state-of-the-art techniques while being
computationally more efficient. We evaluate on prior datasets and additionally
Matterport3D showing qualitative improvements. | [
"cs.CV"
] |
Motion detection in video is important for a number of applications and
fields. In video surveillance, motion detection is an essential accompaniment
to activity recognition for early warning systems. Robotics also has much to
gain from motion detection and segmentation, particularly in high speed motion
tracking for tactile systems. There are a myriad of techniques for detecting
and masking motion in an image. Successful systems have used Gaussian Models to
discern background from foreground in an image (motion from static imagery).
However, particularly in the case of a moving camera or frame of reference, it
is necessary to compensate for the motion of the camera when attempting to
discern objects moving in the foreground. For example, it is possible to
estimate motion of the camera through optical flow methods or temporal
differencing and then compensate for this motion in a background subtraction
model. We selection a method by Yi et al. using Dual-Mode Single Gaussian
Models which does just this. We implement the technique in Intel's Thread
Building Blocks (TBB) and NVIDIA's CUDA libraries. We then compare
parallelization improvements with a theoretical analysis of speedups based on
the characteristics of our selected model and attributes of both TBB and CUDA.
We make our implementation available to the public. | [
"cs.CV",
"cs.DC"
] |
We propose a novel neural network architecture based on dual quaternions
which allow for a compact representation of informations with a main focus on
describing rigid body movements. To cover the dynamic behavior inherent to
rigid body movements, we propose recurrent architectures in the neural network.
To further model the interactions between individual rigid bodies as well as
external inputs efficiently, we incorporate a novel attention mechanism
employing dual quaternion algebra. The introduced architecture is trainable by
means of gradient based algorithms. We apply our approach to a parcel
prediction problem where a rigid body with an initial position, orientation,
velocity and angular velocity moves through a fixed simulation environment
which exhibits rich interactions between the parcel and the boundaries. | [
"cs.LG"
] |
Deep Reinforcement Learning (DRL) has recently achieved significant advances
in various domains. However, explaining the policy of RL agents still remains
an open problem due to several factors, one being the complexity of explaining
neural networks decisions. Recently, a group of works have used
decision-tree-based models to learn explainable policies. Soft decision trees
(SDTs) and discretized differentiable decision trees (DDTs) have been
demonstrated to achieve both good performance and share the benefit of having
explainable policies. In this work, we further improve the results for
tree-based explainable RL in both performance and explainability. Our proposal,
Cascading Decision Trees (CDTs) apply representation learning on the decision
path to allow richer expressivity. Empirical results show that in both
situations, where CDTs are used as policy function approximators or as
imitation learners to explain black-box policies, CDTs can achieve better
performances with more succinct and explainable models than SDTs. As a second
contribution our study reveals limitations of explaining black-box policies via
imitation learning with tree-based explainable models, due to its inherent
instability. | [
"cs.LG"
] |
A major challenge in the pharmaceutical industry is to design novel molecules
with specific desired properties, especially when the property evaluation is
costly. Here, we propose MNCE-RL, a graph convolutional policy network for
molecular optimization with molecular neighborhood-controlled embedding
grammars through reinforcement learning. We extend the original
neighborhood-controlled embedding grammars to make them applicable to molecular
graph generation and design an efficient algorithm to infer grammatical
production rules from given molecules. The use of grammars guarantees the
validity of the generated molecular structures. By transforming molecular
graphs to parse trees with the inferred grammars, the molecular structure
generation task is modeled as a Markov decision process where a policy gradient
strategy is utilized. In a series of experiments, we demonstrate that our
approach achieves state-of-the-art performance in a diverse range of molecular
optimization tasks and exhibits significant superiority in optimizing molecular
properties with a limited number of property evaluations. | [
"cs.LG",
"q-bio.BM"
] |
Interactions between users and videos are the major data source of performing
video recommendation. Despite lots of existing recommendation methods, user
behaviors on videos, which imply the complex relations between users and
videos, are still far from being fully explored. In the paper, we present a
model named Sagittarius. Sagittarius adopts a graph convolutional neural
network to capture the influence between users and videos. In particular,
Sagittarius differentiates between different user behaviors by weighting and
fuses the semantics of user behaviors into the embeddings of users and videos.
Moreover, Sagittarius combines multiple optimization objectives to learn user
and video embeddings and then achieves the video recommendation by the learned
user and video embeddings. The experimental results on multiple datasets show
that Sagittarius outperforms several state-of-the-art models in terms of
recall, unique recall and NDCG. | [
"cs.CV"
] |
Trajectory prediction is critical for applications of planning safe future
movements and remains challenging even for the next few seconds in urban mixed
traffic. How an agent moves is affected by the various behaviors of its
neighboring agents in different environments. To predict movements, we propose
an end-to-end generative model named Attentive Maps Encoder Network (AMENet)
that encodes the agent's motion and interaction information for accurate and
realistic multi-path trajectory prediction. A conditional variational
auto-encoder module is trained to learn the latent space of possible future
paths based on attentive dynamic maps for interaction modeling and then is used
to predict multiple plausible future trajectories conditioned on the observed
past trajectories. The efficacy of AMENet is validated using two public
trajectory prediction benchmarks Trajnet and InD. | [
"cs.CV"
] |
The knowledge of a deep learning model may be transferred to a student model,
leading to intellectual property infringement or vulnerability propagation.
Detecting such knowledge reuse is nontrivial because the suspect models may not
be white-box accessible and/or may serve different tasks. In this paper, we
propose ModelDiff, a testing-based approach to deep learning model similarity
comparison. Instead of directly comparing the weights, activations, or outputs
of two models, we compare their behavioral patterns on the same set of test
inputs. Specifically, the behavioral pattern of a model is represented as a
decision distance vector (DDV), in which each element is the distance between
the model's reactions to a pair of inputs. The knowledge similarity between two
models is measured with the cosine similarity between their DDVs. To evaluate
ModelDiff, we created a benchmark that contains 144 pairs of models that cover
most popular model reuse methods, including transfer learning, model
compression, and model stealing. Our method achieved 91.7% correctness on the
benchmark, which demonstrates the effectiveness of using ModelDiff for model
reuse detection. A study on mobile deep learning apps has shown the feasibility
of ModelDiff on real-world models. | [
"cs.LG",
"cs.AI",
"cs.SE"
] |
In this paper, we propose a novel evaluation metric for performance
evaluation of semantic segmentation. In recent years, many studies have tried
to train pixel-level classifiers on large-scale image datasets to perform
accurate semantic segmentation. The goal of semantic segmentation is to assign
a class label of each pixel in the scene. It has various potential applications
in computer vision fields e.g., object detection, classification, scene
understanding and Etc. To validate the proposed wIoU evaluation metric, we
tested state-of-the art methods on public benchmark datasets (e.g., KITTI)
based on the proposed wIoU metric and compared with other conventional
evaluation metrics. | [
"cs.CV"
] |
The Deep Boltzmann Machines (DBM) is a state-of-the-art unsupervised learning
model, which has been successfully applied to handwritten digit recognition
and, as well as object recognition. However, the DBM is limited in scene
recognition due to the fact that natural scene images are usually very large.
In this paper, an efficient scene recognition approach is proposed based on
superpixels and the DBMs. First, a simple linear iterative clustering (SLIC)
algorithm is employed to generate superpixels of input images, where each
superpixel is regarded as an input of a learning model. Then, a two-layer DBM
model is constructed by stacking two restricted Boltzmann machines (RBMs), and
a greedy layer-wise algorithm is applied to train the DBM model. Finally, a
softmax regression is utilized to categorize scene images. The proposed
technique can effectively reduce the computational complexity and enhance the
performance for large natural image recognition. The approach is verified and
evaluated by extensive experiments, including the fifteen-scene categories
dataset the UIUC eight-sports dataset, and the SIFT flow dataset, are used to
evaluate the proposed method. The experimental results show that the proposed
approach outperforms other state-of-the-art methods in terms of recognition
rate. | [
"cs.CV"
] |
Forecasting multivariate time series is challenging as the variables are
intertwined in time and space, like in the case of traffic signals. Defining
signals on graphs relaxes such complexities by representing the evolution of
signals over a space using relevant graph kernels such as the heat diffusion
kernel. However, this kernel alone does not fully capture the actual dynamics
of the data as it only relies on the graph structure. The gap can be filled by
combining the graph kernel representation with data-driven models that utilize
historical data. This paper proposes a traffic propagation model that merges
multiple heat diffusion kernels into a data-driven prediction model to forecast
traffic signals. We optimize the model parameters using Bayesian inference to
minimize the prediction errors and, consequently, determine the mixing ratio of
the two approaches. Such mixing ratio strongly depends on training data size
and data anomalies, which typically correspond to the peak hours for traffic
data. The proposed model demonstrates prediction accuracy comparable to that of
the state-of-the-art deep neural networks with lower computational effort. It
particularly shows excellent performance for long-term prediction since it
inherits the data-driven models' periodicity modeling. | [
"cs.LG"
] |
Blind image deblurring is a fundamental and challenging computer vision
problem, which aims to recover both the blur kernel and the latent sharp image
from only a blurry observation. Despite the superiority of deep learning
methods in image deblurring have displayed, there still exists major challenge
with various non-uniform motion blur. Previous methods simply take all the
image features as the input to the decoder, which handles different degrees
(e.g. large blur, small blur) simultaneously, leading to challenges for sharp
image generation. To tackle the above problems, we present a deep two-branch
network to deal with blurry images via a component divided module, which
divides an image into two components based on the representation of blurry
degree. Specifically, two component attentive blocks are employed to learn
attention maps to exploit useful deblurring feature representations on both
large and small blurry regions. Then, the blur-aware features are fed into
two-branch reconstruction decoders respectively. In addition, a new feature
fusion mechanism, orientation-based feature fusion, is proposed to merge sharp
features of the two branches. Both qualitative and quantitative experimental
results show that our method performs favorably against the state-of-the-art
approaches. | [
"cs.CV"
] |
Inference capabilities of machine learning (ML) systems skyrocketed in recent
years, now playing a pivotal role in various aspect of society. The goal in
statistical learning is to use data to obtain simple algorithms for predicting
a random variable $Y$ from a correlated observation $X$. Since the dimension of
$X$ is typically huge, computationally feasible solutions should summarize it
into a lower-dimensional feature vector $T$, from which $Y$ is predicted. The
algorithm will successfully make the prediction if $T$ is a good proxy of $Y$,
despite the said dimensionality-reduction. A myriad of ML algorithms (mostly
employing deep learning (DL)) for finding such representations $T$ based on
real-world data are now available. While these methods are often effective in
practice, their success is hindered by the lack of a comprehensive theory to
explain it. The information bottleneck (IB) theory recently emerged as a bold
information-theoretic paradigm for analyzing DL systems. Adopting mutual
information as the figure of merit, it suggests that the best representation
$T$ should be maximally informative about $Y$ while minimizing the mutual
information with $X$. In this tutorial we survey the information-theoretic
origins of this abstract principle, and its recent impact on DL. For the
latter, we cover implications of the IB problem on DL theory, as well as
practical algorithms inspired by it. Our goal is to provide a unified and
cohesive description. A clear view of current knowledge is particularly
important for further leveraging IB and other information-theoretic ideas to
study DL models. | [
"cs.LG",
"stat.ML"
] |
The challenge of learning disentangled representation has recently attracted
much attention and boils down to a competition using a new real world
disentanglement dataset (Gondal et al., 2019). Various methods based on
variational auto-encoder have been proposed to solve this problem, by enforcing
the independence between the representation and modifying the regularization
term in the variational lower bound. However recent work by Locatello et al.
(2018) has demonstrated that the proposed methods are heavily influenced by
randomness and the choice of the hyper-parameter. In this work, instead of
designing a new regularization term, we adopt the FactorVAE but improve the
reconstruction performance and increase the capacity of network and the
training step. The strategy turns out to be very effective and achieve the 1st
place in the challenge. | [
"cs.LG",
"stat.ML"
] |
The ability to generate complex and realistic human body animations at scale,
while following specific artistic constraints, has been a fundamental goal for
the game and animation industry for decades. Popular techniques include
key-framing, physics-based simulation, and database methods via motion graphs.
Recently, motion generators based on deep learning have been introduced.
Although these learning models can automatically generate highly intricate
stylized motions of arbitrary length, they still lack user control. To this
end, we introduce the problem of long-term inbetweening, which involves
automatically synthesizing complex motions over a long time interval given very
sparse keyframes by users. We identify a number of challenges related to this
problem, including maintaining biomechanical and keyframe constraints,
preserving natural motions, and designing the entire motion sequence
holistically while considering all constraints. We introduce a biomechanically
constrained generative adversarial network that performs long-term inbetweening
of human motions, conditioned on keyframe constraints. This network uses a
novel two-stage approach where it first predicts local motion in the form of
joint angles, and then predicts global motion, i.e. the global path that the
character follows. Since there are typically a number of possible motions that
could satisfy the given user constraints, we also enable our network to
generate a variety of outputs with a scheme that we call Motion DNA. This
approach allows the user to manipulate and influence the output content by
feeding seed motions (DNA) to the network. Trained with 79 classes of captured
motion data, our network performs robustly on a variety of highly complex
motion styles. | [
"cs.CV",
"cs.GR"
] |
We present a multi-relational temporal Knowledge Graph based on the daily
interactions between artifacts in GitHub, one of the largest social coding
platforms. Such representation enables posing many user-activity and project
management questions as link prediction and time queries over the knowledge
graph. In particular, we introduce two new datasets for i) interpolated
time-conditioned link prediction and ii) extrapolated time-conditioned
link/time prediction queries, each with distinguished properties. Our
experiments on these datasets highlight the potential of adapting knowledge
graphs to answer broad software engineering questions. Meanwhile, it also
reveals the unsatisfactory performance of existing temporal models on
extrapolated queries and time prediction queries in general. To overcome these
shortcomings, we introduce an extension to current temporal models using
relative temporal information with regards to past events. | [
"cs.LG",
"cs.SE",
"stat.ML"
] |
We study the problem of end-to-end learning from complex multigraphs with
potentially very large numbers of edges between two vertices, each edge labeled
with rich information. Examples range from communication networks to flights
between airports or financial transaction graphs. We propose Latent-Graph
Convolutional Networks (L-GCNs), which propagate information from these complex
edges to a latent adjacency tensor, after which further downstream tasks can be
performed, such as node classification. We evaluate the performance of several
variations of the model on two synthetic datasets simulating fraud in financial
transaction networks, ensuring the model must make use of edge labels in order
to achieve good classification performance. We find that allowing for nonlinear
interactions on a per-neighbor basis boosts performance significantly, while
showing promising results in an inductive setting. Finally, we demonstrate the
use of L-GCNs on real-world data in the form of an urban transportation
network. | [
"stat.ML",
"cs.LG",
"cs.SI"
] |
Recent works have demonstrated that increasing model capacity through width
in over-parameterized neural networks leads to a decrease in test risk. For
neural networks, however, model capacity can also be increased through depth,
yet understanding the impact of increasing depth on test risk remains an open
question. In this work, we demonstrate that the test risk of over-parameterized
convolutional networks is a U-shaped curve (i.e. monotonically decreasing, then
increasing) with increasing depth. We first provide empirical evidence for this
phenomenon via image classification experiments using both ResNets and the
convolutional neural tangent kernel (CNTK). We then present a novel linear
regression framework for characterizing the impact of depth on test risk, and
show that increasing depth leads to a U-shaped test risk for the linear CNTK.
In particular, we prove that the linear CNTK corresponds to a depth-dependent
linear transformation on the original space and characterize properties of this
transformation. We then analyze over-parameterized linear regression under
arbitrary linear transformations and, in simplified settings, provably identify
the depths which minimize each of the bias and variance terms of the test risk. | [
"cs.LG",
"stat.ML"
] |
We present a novel real-time line segment detection scheme called Line Graph
Neural Network (LGNN). Existing approaches require a computationally expensive
verification or postprocessing step. Our LGNN employs a deep convolutional
neural network (DCNN) for proposing line segment directly, with a graph neural
network (GNN) module for reasoning their connectivities. Specifically, LGNN
exploits a new quadruplet representation for each line segment where the GNN
module takes the predicted candidates as vertexes and constructs a sparse graph
to enforce structural context. Compared with the state-of-the-art, LGNN
achieves near real-time performance without compromising accuracy. LGNN further
enables time-sensitive 3D applications. When a 3D point cloud is accessible, we
present a multi-modal line segment classification technique for extracting a 3D
wireframe of the environment robustly and efficiently. | [
"cs.CV",
"cs.LG",
"eess.IV"
] |
Recommender systems today have become an essential component of any
commercial website. Collaborative filtering approaches, and Matrix
Factorization (MF) techniques in particular, are widely used in recommender
systems. However, the natural data sparsity problem limits their performance
where users generally interact with very few items in the system. Consequently,
multiple hybrid models were proposed recently to optimize MF performance by
incorporating additional contextual information in its learning process.
Although these models improve the recommendation quality, there are two primary
aspects for further improvements: (1) multiple models focus only on some
portion of the available contextual information and neglect other portions; (2)
learning the feature space of the side contextual information needs to be
further enhanced. In this paper, we introduce a Collaborative Dual Attentive
Autoencoder (CATA++) for recommending scientific articles. CATA++ utilizes an
article's content and learns its latent space via two parallel autoencoders. We
employ the attention mechanism to capture the most related parts of information
in order to make more relevant recommendations. Extensive experiments on three
real-world datasets have shown that our dual-way learning strategy has
significantly improved the MF performance in comparison with other
state-of-the-art MF-based models using various experimental evaluations. The
source code of our methods is available at:
https://github.com/jianlin-cheng/CATA. | [
"cs.LG",
"cs.IR",
"stat.ML"
] |
In the last decade, supervised deep learning approaches have been extensively
employed in visual odometry (VO) applications, which is not feasible in
environments where labelled data is not abundant. On the other hand,
unsupervised deep learning approaches for localization and mapping in unknown
environments from unlabelled data have received comparatively less attention in
VO research. In this study, we propose a generative unsupervised learning
framework that predicts 6-DoF pose camera motion and monocular depth map of the
scene from unlabelled RGB image sequences, using deep convolutional Generative
Adversarial Networks (GANs). We create a supervisory signal by warping view
sequences and assigning the re-projection minimization to the objective loss
function that is adopted in multi-view pose estimation and single-view depth
generation network. Detailed quantitative and qualitative evaluations of the
proposed framework on the KITTI and Cityscapes datasets show that the proposed
method outperforms both existing traditional and unsupervised deep VO methods
providing better results for both pose estimation and depth recovery. | [
"cs.LG",
"cs.CV",
"stat.ML"
] |
Having a perfect model to compute the optimal policy is often infeasible in
reinforcement learning. It is important in high-stakes domains to quantify and
manage risk induced by model uncertainties. Entropic risk measure is an
exponential utility-based convex risk measure that satisfies many reasonable
properties. In this paper, we propose an entropic risk constrained policy
gradient and actor-critic algorithms that are risk-averse to the model
uncertainty. We demonstrate the usefulness of our algorithms on several problem
domains. | [
"cs.LG",
"math.OC",
"stat.ML"
] |
Detection of anomalous trajectories is an important problem with potential
applications to various domains, such as video surveillance, risk assessment,
vessel monitoring and high-energy physics. Modeling the distribution of
trajectories with statistical approaches has been a challenging task due to the
fact that such time series are usually non stationary and highly dimensional.
However, modern machine learning techniques provide robust approaches for
data-driven modeling and critical information extraction. In this paper, we
propose a Sequence to Sequence architecture for real-time detection of
anomalies in human trajectories, in the context of risk-based security. Our
detection scheme is tested on a synthetic dataset of diverse and realistic
trajectories generated by the ISL iCrowd simulator. The experimental results
indicate that our scheme accurately detects motion patterns that deviate from
normal behaviors and is promising for future real-world applications. | [
"cs.LG",
"cs.CV",
"eess.IV"
] |
Drug combination therapy has become a increasingly promising method in the
treatment of cancer. However, the number of possible drug combinations is so
huge that it is hard to screen synergistic drug combinations through wet-lab
experiments. Therefore, computational screening has become an important way to
prioritize drug combinations. Graph neural network have recently shown
remarkable performance in the prediction of compound-protein interactions, but
it has not been applied to the screening of drug combinations. In this paper,
we proposed a deep learning model based on graph neural networks and attention
mechanism to identify drug combinations that can effectively inhibit the
viability of specific cancer cells. The feature embeddings of drug molecule
structure and gene expression profiles were taken as input to multi-layer
feedforward neural network to identify the synergistic drug combinations. We
compared DeepDDS with classical machine learning methods and other deep
learning-based methods on benchmark data set, and the leave-one-out
experimental results showed that DeepDDS achieved better performance than
competitive methods. Also, on an independent test set released by well-known
pharmaceutical enterprise AstraZeneca, DeepDDS was superior to competitive
methods by more than 16\% predictive precision. Furthermore, we explored the
interpretability of the graph attention network, and found the correlation
matrix of atomic features revealed important chemical substructures of drugs.
We believed that DeepDDS is an effective tool that prioritized synergistic drug
combinations for further wet-lab experiment validation. | [
"cs.LG",
"q-bio.QM"
] |
Depth perception is considered an invaluable source of information for
various vision tasks. However, depth maps acquired using consumer-level sensors
still suffer from non-negligible noise. This fact has recently motivated
researchers to exploit traditional filters, as well as the deep learning
paradigm, in order to suppress the aforementioned non-uniform noise, while
preserving geometric details. Despite the effort, deep depth denoising is still
an open challenge mainly due to the lack of clean data that could be used as
ground truth. In this paper, we propose a fully convolutional deep autoencoder
that learns to denoise depth maps, surpassing the lack of ground truth data.
Specifically, the proposed autoencoder exploits multiple views of the same
scene from different points of view in order to learn to suppress noise in a
self-supervised end-to-end manner using depth and color information during
training, yet only depth during inference. To enforce selfsupervision, we
leverage a differentiable rendering technique to exploit photometric
supervision, which is further regularized using geometric and surface priors.
As the proposed approach relies on raw data acquisition, a large RGB-D corpus
is collected using Intel RealSense sensors. Complementary to a quantitative
evaluation, we demonstrate the effectiveness of the proposed self-supervised
denoising approach on established 3D reconstruction applications. Code is
avalable at https://github.com/VCL3D/DeepDepthDenoising | [
"cs.CV"
] |
In just three years, Variational Autoencoders (VAEs) have emerged as one of
the most popular approaches to unsupervised learning of complicated
distributions. VAEs are appealing because they are built on top of standard
function approximators (neural networks), and can be trained with stochastic
gradient descent. VAEs have already shown promise in generating many kinds of
complicated data, including handwritten digits, faces, house numbers, CIFAR
images, physical models of scenes, segmentation, and predicting the future from
static images. This tutorial introduces the intuitions behind VAEs, explains
the mathematics behind them, and describes some empirical behavior. No prior
knowledge of variational Bayesian methods is assumed. | [
"stat.ML",
"cs.LG"
] |
In this paper we introduce a novel method for general semantic segmentation
that can benefit from general semantics of Convolutional Neural Network (CNN).
Our segmentation proposes visually and semantically coherent image segments. We
use binary encoding of CNN features to overcome the difficulty of the
clustering on the high-dimensional CNN feature space. These binary codes are
very robust against noise and non-semantic changes in the image. These binary
encoding can be embedded into the CNN as an extra layer at the end of the
network. This results in real-time segmentation. To the best of our knowledge
our method is the first attempt on general semantic image segmentation using
CNN. All the previous papers were limited to few number of category of the
images (e.g. PASCAL VOC). Experiments show that our segmentation algorithm
outperform the state-of-the-art non-semantic segmentation methods by large
margin. | [
"cs.CV"
] |
We study reinforcement learning (RL) with linear function approximation.
Existing algorithms for this problem only have high-probability regret and/or
Probably Approximately Correct (PAC) sample complexity guarantees, which cannot
guarantee the convergence to the optimal policy. In this paper, in order to
overcome the limitation of existing algorithms, we propose a new algorithm
called FLUTE, which enjoys uniform-PAC convergence to the optimal policy with
high probability. The uniform-PAC guarantee is the strongest possible guarantee
for reinforcement learning in the literature, which can directly imply both PAC
and high probability regret bounds, making our algorithm superior to all
existing algorithms with linear function approximation. At the core of our
algorithm is a novel minimax value function estimator and a multi-level
partition scheme to select the training samples from historical observations.
Both of these techniques are new and of independent interest. | [
"cs.LG",
"math.OC",
"stat.ML"
] |
This paper addresses a major flaw of the cycle consistency loss when used to
preserve the input appearance in the face-to-face synthesis domain. In
particular, we show that the images generated by a network trained using this
loss conceal a noise that hinders their use for further tasks. To overcome this
limitation, we propose a ''recurrent cycle consistency loss" which for
different sequences of target attributes minimises the distance between the
output images, independent of any intermediate step. We empirically validate
not only that our loss enables the re-use of generated images, but that it also
improves their quality. In addition, we propose the very first network that
covers the task of unconstrained landmark-guided face-to-face synthesis.
Contrary to previous works, our proposed approach enables the transfer of a
particular set of input features to a large span of poses and expressions,
whereby the target landmarks become the ground-truth points. We then evaluate
the consistency of our proposed approach to synthesise faces at the target
landmarks. To the best of our knowledge, we are the first to propose a loss to
overcome the limitation of the cycle consistency loss, and the first to propose
an ''in-the-wild'' landmark guided synthesis approach. Code and models for this
paper can be found in https://github.com/ESanchezLozano/GANnotation | [
"cs.CV"
] |
In this work, we propose a fast superpixel-based color transfer method (SCT)
between two images. Superpixels enable to decrease the image dimension and to
extract a reduced set of color candidates. We propose to use a fast approximate
nearest neighbor matching algorithm in which we enforce the match diversity by
limiting the selection of the same superpixels. A fusion framework is designed
to transfer the matched colors, and we demonstrate the improvement obtained
over exact matching results. Finally, we show that SCT is visually competitive
compared to state-of-the-art methods. | [
"cs.CV"
] |
Makeup transfer is the task of applying on a source face the makeup style
from a reference image. Real-life makeups are diverse and wild, which cover not
only color-changing but also patterns, such as stickers, blushes, and
jewelries. However, existing works overlooked the latter components and
confined makeup transfer to color manipulation, focusing only on light makeup
styles. In this work, we propose a holistic makeup transfer framework that can
handle all the mentioned makeup components. It consists of an improved color
transfer branch and a novel pattern transfer branch to learn all makeup
properties, including color, shape, texture, and location. To train and
evaluate such a system, we also introduce new makeup datasets for real and
synthetic extreme makeup. Experimental results show that our framework achieves
the state of the art performance on both light and extreme makeup styles. Code
is available at https://github.com/VinAIResearch/CPM. | [
"cs.CV"
] |
Image to image translation aims to learn a mapping that transforms an image
from one visual domain to another. Recent works assume that images descriptors
can be disentangled into a domain-invariant content representation and a
domain-specific style representation. Thus, translation models seek to preserve
the content of source images while changing the style to a target visual
domain. However, synthesizing new images is extremely challenging especially in
multi-domain translations, as the network has to compose content and style to
generate reliable and diverse images in multiple domains. In this paper we
propose the use of an image retrieval system to assist the image-to-image
translation task. First, we train an image-to-image translation model to map
images to multiple domains. Then, we train an image retrieval model using real
and generated images to find images similar to a query one in content but in a
different domain. Finally, we exploit the image retrieval system to fine-tune
the image-to-image translation model and generate higher quality images. Our
experiments show the effectiveness of the proposed solution and highlight the
contribution of the retrieval network, which can benefit from additional
unlabeled data and help image-to-image translation models in the presence of
scarce data. | [
"cs.CV"
] |
In this work, we construct a large-scale dataset for vehicle
re-identification (ReID), which contains 137k images of 13k vehicle instances
captured by UAV-mounted cameras. To our knowledge, it is the largest UAV-based
vehicle ReID dataset. To increase intra-class variation, each vehicle is
captured by at least two UAVs at different locations, with diverse view-angles
and flight-altitudes. We manually label a variety of vehicle attributes,
including vehicle type, color, skylight, bumper, spare tire and luggage rack.
Furthermore, for each vehicle image, the annotator is also required to mark the
discriminative parts that helps them to distinguish this particular vehicle
from others. Besides the dataset, we also design a specific vehicle ReID
algorithm to make full use of the rich annotation information. It is capable of
explicitly detecting discriminative parts for each specific vehicle and
significantly outperforms the evaluated baselines and state-of-the-art vehicle
ReID approaches. | [
"cs.CV"
] |
After DETR was proposed, this novel transformer-based detection paradigm
which performs several cross-attentions between object queries and feature maps
for predictions has subsequently derived a series of transformer-based
detection heads. These models iterate object queries after each
cross-attention. However, they don't renew the query position which indicates
object queries' position information. Thus model needs extra learning to figure
out the newest regions that query position should express and need more
attention. To fix this issue, we propose the Guided Query Position (GQPos)
method to embed the latest location information of object queries to query
position iteratively.
Another problem of such transformer-based detection heads is the high
complexity to perform attention on multi-scale feature maps, which hinders them
from improving detection performance at all scales. Therefore we propose a
novel fusion scheme named Similar Attention (SiA): besides the feature maps is
fused, SiA also fuse the attention weights maps to accelerate the learning of
high-resolution attention weight map by well-learned low-resolution attention
weight map.
Our experiments show that the proposed GQPos improves the performance of a
series of models, including DETR, SMCA, YoloS, and HoiTransformer and SiA
consistently improve the performance of multi-scale transformer-based detection
heads like DETR and HoiTransformer. | [
"cs.CV"
] |
In the last decade, many diverse advances have occurred in the field of
information extraction from data. Information extraction in its simplest form
takes place in computing environments, where structured data can be extracted
through a series of queries. The continuous expansion of quantities of data
have therefore provided an opportunity for knowledge extraction (KE) from a
textual document (TD). A typical problem of this kind is the extraction of
common characteristics and knowledge from a group of TDs, with the possibility
to group such similar TDs in a process known as clustering. In this paper we
present a technique for such KE among a group of TDs related to the common
characteristics and meaning of their content. Our technique is based on the
Spearman's Rank Correlation Coefficient (SRCC), for which the conducted
experiments have proven to be comprehensive measure to achieve a high-quality
KE. | [
"cs.LG",
"cs.CL",
"cs.IR",
"stat.ML"
] |
This paper focuses on a class of reinforcement learning problems where
significant events are rare and limited to a single positive reward per
episode. A typical example is that of an agent who has to choose a partner to
cooperate with, while a large number of partners are simply not interested in
cooperating, regardless of what the agent has to offer. We address this problem
in a continuous state and action space with two different kinds of search
methods: a gradient policy search method and a direct policy search method
using an evolution strategy. We show that when significant events are rare,
gradient information is also scarce, making it difficult for policy gradient
search methods to find an optimal policy, with or without a deep neural
architecture. On the other hand, we show that direct policy search methods are
invariant to the rarity of significant events, which is yet another
confirmation of the unique role evolutionary algorithms has to play as a
reinforcement learning method. | [
"cs.LG",
"cs.AI",
"cs.NE",
"I.2.6; I.2"
] |
Learning rich and diverse representations is critical for the performance of
deep convolutional neural networks (CNNs). In this paper, we consider how to
use privileged information to promote inherent diversity of a single CNN model
such that the model can learn better representations and offer stronger
generalization ability. To this end, we propose a novel group orthogonal
convolutional neural network (GoCNN) that learns untangled representations
within each layer by exploiting provided privileged information and enhances
representation diversity effectively. We take image classification as an
example where image segmentation annotations are used as privileged information
during the training process. Experiments on two benchmark datasets -- ImageNet
and PASCAL VOC -- clearly demonstrate the strong generalization ability of our
proposed GoCNN model. On the ImageNet dataset, GoCNN improves the performance
of state-of-the-art ResNet-152 model by absolute value of 1.2% while only uses
privileged information of 10% of the training images, confirming effectiveness
of GoCNN on utilizing available privileged knowledge to train better CNNs. | [
"cs.CV"
] |
Object detection is a fundamental and challenging problem in aerial and
satellite image analysis. More recently, a two-stage detector Faster R-CNN is
proposed and demonstrated to be a promising tool for object detection in
optical remote sensing images, while the sparse and dense characteristic of
objects in remote sensing images is complexity. It is unreasonable to treat all
images with the same region proposal strategy, and this treatment limits the
performance of two-stage detectors. In this paper, we propose a novel and
effective approach, named deep adaptive proposal network (DAPNet), address this
complexity characteristic of object by learning a new category prior network
(CPN) on the basis of the existing Faster R-CNN architecture. Moreover, the
candidate regions produced by DAPNet model are different from the traditional
region proposal network (RPN), DAPNet predicts the detail category of each
candidate region. And these candidate regions combine the object number, which
generated by the category prior network to achieve a suitable number of
candidate boxes for each image. These candidate boxes can satisfy detection
tasks in sparse and dense scenes. The performance of the proposed framework has
been evaluated on the challenging NWPU VHR-10 data set. Experimental results
demonstrate the superiority of the proposed framework to the state-of-the-art. | [
"cs.CV"
] |
This paper addresses fast semantic segmentation on video.Video segmentation
often calls for real-time, or even fasterthan real-time, processing. One common
recipe for conserving computation arising from feature extraction is to
propagate features of few selected keyframes. However, recent advances in fast
image segmentation make these solutions less attractive. To leverage fast image
segmentation for furthering video segmentation, we propose a simple yet
efficient propagation framework. Specifically, we perform lightweight flow
estimation in 1/8-downscaled image space for temporal warping in segmentation
outpace space. Moreover, we introduce a guided spatially-varying convolution
for fusing segmentations derived from the previous and current frames, to
mitigate propagation error and enable lightweight feature extraction on
non-keyframes. Experimental results on Cityscapes and CamVid show that our
scheme achieves the state-of-the-art accuracy-throughput trade-off on video
segmentation. | [
"cs.CV",
"cs.LG"
] |
Attention mechanisms, especially self-attention, have played an increasingly
important role in deep feature representation for visual tasks. Self-attention
updates the feature at each position by computing a weighted sum of features
using pair-wise affinities across all positions to capture the long-range
dependency within a single sample. However, self-attention has quadratic
complexity and ignores potential correlation between different samples. This
paper proposes a novel attention mechanism which we call external attention,
based on two external, small, learnable, shared memories, which can be
implemented easily by simply using two cascaded linear layers and two
normalization layers; it conveniently replaces self-attention in existing
popular architectures. External attention has linear complexity and implicitly
considers the correlations between all data samples. We further incorporate the
multi-head mechanism into external attention to provide an all-MLP
architecture, external attention MLP (EAMLP), for image classification.
Extensive experiments on image classification, object detection, semantic
segmentation, instance segmentation, image generation, and point cloud analysis
reveal that our method provides results comparable or superior to the
self-attention mechanism and some of its variants, with much lower
computational and memory costs. | [
"cs.CV"
] |
Medical image datasets are usually imbalanced, due to the high costs of
obtaining the data and time-consuming annotations. Training deep neural network
models on such datasets to accurately classify the medical condition does not
yield desired results and often over-fits the data on majority class samples.
In order to address this issue, data augmentation is often performed on
training data by position augmentation techniques such as scaling, cropping,
flipping, padding, rotation, translation, affine transformation, and color
augmentation techniques such as brightness, contrast, saturation, and hue to
increase the dataset sizes. These augmentation techniques are not guaranteed to
be advantageous in domains with limited data, especially medical image data,
and could lead to further overfitting. In this work, we performed data
augmentation on the Chest X-rays dataset through generative modeling (deep
convolutional generative adversarial network) which creates artificial
instances retaining similar characteristics to the original data and evaluation
of the model resulted in Fr\'echet Distance of Inception (FID) score of 1.289. | [
"cs.CV",
"cs.AI",
"cs.LG"
] |
Sample efficiency and risk-awareness are central to the development of
practical reinforcement learning (RL) for complex decision-making. The former
can be addressed by transfer learning and the latter by optimizing some utility
function of the return. However, the problem of transferring skills in a
risk-aware manner is not well-understood. In this paper, we address the problem
of risk-aware policy transfer between tasks in a common domain that differ only
in their reward functions, in which risk is measured by the variance of reward
streams. Our approach begins by extending the idea of generalized policy
improvement to maximize entropic utilities, thus extending policy improvement
via dynamic programming to sets of policies and levels of risk-aversion. Next,
we extend the idea of successor features (SF), a value function representation
that decouples the environment dynamics from the rewards, to capture the
variance of returns. Our resulting risk-aware successor features (RaSF)
integrate seamlessly within the RL framework, inherit the superior task
generalization ability of SFs, and incorporate risk-awareness into the
decision-making. Experiments on a discrete navigation domain and control of a
simulated robotic arm demonstrate the ability of RaSFs to outperform
alternative methods including SFs, when taking the risk of the learned policies
into account. | [
"cs.LG",
"cs.AI",
"cs.RO"
] |
Recently, the Weisfeiler-Lehman (WL) graph isomorphism test was used to
measure the expressiveness of graph neural networks (GNNs), showing that the
neighborhood aggregation GNNs were at most as powerful as 1-WL test in
distinguishing graph structures. There were also improvements proposed in
analogy to $k$-WL test ($k>1$). However, the aggregators in these GNNs are far
from injective as required by the WL test, and suffer from weak distinguishing
strength, making it become expressive bottlenecks. In this paper, we improve
the expressiveness by exploring powerful aggregators. We reformulate
aggregation with the corresponding aggregation coefficient matrix, and then
systematically analyze the requirements of the aggregation coefficient matrix
for building more powerful aggregators and even injective aggregators. It can
also be viewed as the strategy for preserving the rank of hidden features, and
implies that basic aggregators correspond to a special case of low-rank
transformations. We also show the necessity of applying nonlinear units ahead
of aggregation, which is different from most aggregation-based GNNs. Based on
our theoretical analysis, we develop two GNN layers, ExpandingConv and
CombConv. Experimental results show that our models significantly boost
performance, especially for large and densely connected graphs. | [
"cs.LG",
"cs.AI"
] |
Softmax working with cross-entropy is widely used in classification, which
evaluates the similarity between two discrete distribution columns (predictions
and true labels). Inspired by chi-square test, we designed a new loss function
called chi-square loss, which is also works for Softmax. Chi-square loss has a
statistical background. We proved that it is unbiased in optimization, and
clarified its using conditions (its formula determines that it must work with
label smoothing). In addition, we studied the sample distribution of this loss
function by visualization and found that the distribution is related to the
neural network structure, which is distinct compared to cross-entropy. In the
past, the influence of structure was often ignored when visualizing. Chi-square
loss can notice changes in neural network structure because it is very strict,
and we explained the reason for this strictness. We also studied the influence
of label smoothing and discussed the relationship between label smoothing and
training accuracy and stability. Since the chi-square loss is very strict, the
performance will degrade when dealing samples of very many classes. | [
"cs.LG",
"cs.AI"
] |
Predicting the properties of a molecule from its structure is a challenging
task. Recently, deep learning methods have improved the state of the art for
this task because of their ability to learn useful features from the given
data. By treating molecule structure as graphs, where atoms and bonds are
modeled as nodes and edges, graph neural networks (GNNs) have been widely used
to predict molecular properties. However, the design and development of GNNs
for a given data set rely on labor-intensive design and tuning of the network
architectures. Neural architecture search (NAS) is a promising approach to
discover high-performing neural network architectures automatically. To that
end, we develop an NAS approach to automate the design and development of GNNs
for molecular property prediction. Specifically, we focus on automated
development of message-passing neural networks (MPNNs) to predict the molecular
properties of small molecules in quantum mechanics and physical chemistry data
sets from the MoleculeNet benchmark. We demonstrate the superiority of the
automatically discovered MPNNs by comparing them with manually designed GNNs
from the MoleculeNet benchmark. We study the relative importance of the choices
in the MPNN search space, demonstrating that customizing the architecture is
critical to enhancing performance in molecular property prediction and that the
proposed approach can perform customization automatically with minimal manual
effort. | [
"cs.LG",
"q-bio.BM",
"stat.ML"
] |
Recent research on reinforcement learning (RL) has suggested that trained
agents are vulnerable to maliciously crafted adversarial samples. In this work,
we show how such samples can be generalised from White-box and Grey-box attacks
to a strong Black-box case, where the attacker has no knowledge of the agents,
their training parameters and their training methods. We use
sequence-to-sequence models to predict a single action or a sequence of future
actions that a trained agent will make. First, we show our approximation model,
based on time-series information from the agent, consistently predicts RL
agents' future actions with high accuracy in a Black-box setup on a wide range
of games and RL algorithms. Second, we find that although adversarial samples
are transferable from the target model to our RL agents, they often outperform
random Gaussian noise only marginally. This highlights a serious methodological
deficiency in previous work on such agents; random jamming should have been
taken as the baseline for evaluation. Third, we propose a novel use for
adversarial samplesin Black-box attacks of RL agents: they can be used to
trigger a trained agent to misbehave after a specific time delay. This appears
to be a genuinely new type of attack. It potentially enables an attacker to use
devices controlled by RL agents as time bombs. | [
"cs.LG",
"cs.CR",
"cs.CV",
"stat.ML"
] |
Variance plays a crucial role in risk-sensitive reinforcement learning, and
most risk measures can be analyzed via variance. In this paper, we consider two
law-invariant risks as examples: mean-variance risk and exponential utility
risk. With the aid of the state-augmentation transformation (SAT), we show
that, the two risks can be estimated in Markov decision processes (MDPs) with a
stochastic transition-based reward and a randomized policy. To relieve the
enlarged state space, a novel definition of isotopic states is proposed for
state lumping, considering the special structure of the transformed transition
probability. In the numerical experiment, we illustrate state lumping in the
SAT, errors from a naive reward simplification, and the validity of the SAT for
the two risk estimations. | [
"cs.LG",
"cs.AI",
"stat.ML"
] |
Towards better unsupervised domain adaptation (UDA). Recently, researchers
propose various domain-conditioned attention modules and make promising
progresses. However, considering that the configuration of attention, i.e., the
type and the position of attention module, affects the performance
significantly, it is more generalized to optimize the attention configuration
automatically to be specialized for arbitrary UDA scenario. For the first time,
this paper proposes EvoADA: a novel framework to evolve the attention
configuration for a given UDA task without human intervention. In particular,
we propose a novel search space containing diverse attention configurations.
Then, to evaluate the attention configurations and make search procedure
UDA-oriented (transferability + discrimination), we apply a simple and
effective evaluation strategy: 1) training the network weights on two domains
with off-the-shelf domain adaptation methods; 2) evolving the attention
configurations under the guide of the discriminative ability on the target
domain. Experiments on various kinds of cross-domain benchmarks, i.e.,
Office-31, Office-Home, CUB-Paintings, and Duke-Market-1510, reveal that the
proposed EvoADA consistently boosts multiple state-of-the-art domain adaptation
approaches, and the optimal attention configurations help them achieve better
performance. | [
"cs.CV"
] |
Deep generative models have recently been applied to molecule design. If the
molecules are encoded in linear SMILES strings, modeling becomes convenient.
However, models relying on string representations tend to generate invalid
samples and duplicates. Prior work addressed these issues by building models on
chemically-valid fragments or explicitly enforcing chemical rules in the
generation process. We argue that an expressive model is sufficient to
implicitly and automatically learn the complicated chemical rules from the
data, even if molecules are encoded in simple character-level SMILES strings.
We propose to learn latent space energy-based prior model with SMILES
representation for molecule modeling. Our experiments show that our method is
able to generate molecules with validity and uniqueness competitive with
state-of-the-art models. Interestingly, generated molecules have structural and
chemical features whose distributions almost perfectly match those of the real
molecules. | [
"cs.LG"
] |
We introduce a novel approach to feed-forward neural network interpretation
based on partitioning the space of sequences of neuron activations. In line
with this approach, we propose a model-specific interpretation method, called
YASENN. Our method inherits many advantages of model-agnostic distillation,
such as an ability to focus on the particular input region and to express an
explanation in terms of features different from those observed by a neural
network. Moreover, examination of distillation error makes the method
applicable to the problems with low tolerance to interpretation mistakes.
Technically, YASENN distills the network with an ensemble of layer-wise
gradient boosting decision trees and encodes the sequences of neuron
activations with leaf indices. The finite number of unique codes induces a
partitioning of the input space. Each partition may be described in a variety
of ways, including examination of an interpretable model (e.g. a logistic
regression or a decision tree) trained to discriminate between objects of those
partitions. Our experiments provide an intuition behind the method and
demonstrate revealed artifacts in neural network decision making. | [
"cs.LG",
"stat.ML"
] |
We report resolution enhancement in scanning electron microscopy (SEM) images
using a generative adversarial network. We demonstrate the veracity of this
deep learning-based super-resolution technique by inferring unresolved features
in low-resolution SEM images and comparing them with the accurately
co-registered high-resolution SEM images of the same samples. Through spatial
frequency analysis, we also report that our method generates images with
frequency spectra matching higher resolution SEM images of the same
fields-of-view. By using this technique, higher resolution SEM images can be
taken faster, while also reducing both electron charging and damage to the
samples. | [
"cs.CV",
"cs.LG",
"physics.app-ph"
] |
Joint object detection and semantic segmentation can be applied to many
fields, such as self-driving cars and unmanned surface vessels. An initial and
important progress towards this goal has been achieved by simply sharing the
deep convolutional features for the two tasks. However, this simple scheme is
unable to make full use of the fact that detection and segmentation are
mutually beneficial. To overcome this drawback, we propose a framework called
TripleNet where triple supervisions including detection-oriented supervision,
class-aware segmentation supervision, and class-agnostic segmentation
supervision are imposed on each layer of the decoder network. Class-agnostic
segmentation supervision provides an objectness prior knowledge for both
semantic segmentation and object detection. Besides the three types of
supervisions, two light-weight modules (i.e., inner-connected module and
attention skip-layer fusion) are also incorporated into each layer of the
decoder. In the proposed framework, detection and segmentation can sufficiently
boost each other. Moreover, class-agnostic and class-aware segmentation on each
decoder layer are not performed at the test stage. Therefore, no extra
computational costs are introduced at the test stage. Experimental results on
the VOC2007 and VOC2012 datasets demonstrate that the proposed TripleNet is
able to improve both the detection and segmentation accuracies without adding
extra computational costs. | [
"cs.CV"
] |
Being able to reach any desired location in the environment can be a valuable
asset for an agent. Learning a policy to navigate between all pairs of states
individually is often not feasible. An all-goals updating algorithm uses each
transition to learn Q-values towards all goals simultaneously and off-policy.
However the expensive numerous updates in parallel limited the approach to
small tabular cases so far. To tackle this problem we propose to use
convolutional network architectures to generate Q-values and updates for a
large number of goals at once. We demonstrate the accuracy and generalization
qualities of the proposed method on randomly generated mazes and Sokoban
puzzles. In the case of on-screen goal coordinates the resulting mapping from
frames to distance-maps directly informs the agent about which places are
reachable and in how many steps. As an example of application we show that
replacing the random actions in epsilon-greedy exploration by several actions
towards feasible goals generates better exploratory trajectories on Montezuma's
Revenge and Super Mario All-Stars games. | [
"cs.LG",
"stat.ML"
] |
Feature interactions across space and scales underpin modern visual
recognition systems because they introduce beneficial visual contexts.
Conventionally, spatial contexts are passively hidden in the CNN's increasing
receptive fields or actively encoded by non-local convolution. Yet, the
non-local spatial interactions are not across scales, and thus they fail to
capture the non-local contexts of objects (or parts) residing in different
scales. To this end, we propose a fully active feature interaction across both
space and scales, called Feature Pyramid Transformer (FPT). It transforms any
feature pyramid into another feature pyramid of the same size but with richer
contexts, by using three specially designed transformers in self-level,
top-down, and bottom-up interaction fashion. FPT serves as a generic visual
backbone with fair computational overhead. We conduct extensive experiments in
both instance-level (i.e., object detection and instance segmentation) and
pixel-level segmentation tasks, using various backbones and head networks, and
observe consistent improvement over all the baselines and the state-of-the-art
methods. | [
"cs.CV"
] |
Current deep domain adaptation methods used in computer vision have mainly
focused on learning discriminative and domain-invariant features across
different domains. In this paper, we present a novel "deep adversarial
transition learning" (DATL) framework that bridges the domain gap by projecting
the source and target domains into intermediate, transitional spaces through
the employment of adjustable, cross-grafted generative network stacks and
effective adversarial learning between transitions. Specifically, we construct
variational auto-encoders (VAE) for the two domains, and form bidirectional
transitions by cross-grafting the VAEs' decoder stacks. Furthermore, generative
adversarial networks (GAN) are employed for domain adaptation, mapping the
target domain data to the known label space of the source domain. The overall
adaptation process hence consists of three phases: feature representation
learning by VAEs, transitions generation, and transitions alignment by GANs.
Experimental results demonstrate that our method outperforms the state-of-the
art on a number of unsupervised domain adaptation benchmarks. | [
"cs.CV",
"cs.LG",
"I.2.10; I.2.6"
] |
The image nonlocal self-similarity (NSS) prior refers to the fact that a
local patch often has many nonlocal similar patches to it across the image. In
this paper we apply such NSS prior to enhance the robust quaternion matrix
completion (QMC) method and significantly improve the inpainting performance. A
patch group based NSS prior learning scheme is proposed to learn explicit NSS
models from natural color images. The NSS-based QMC algorithm computes an
optimal low-rank approximation to the high-rank color image, resulting in high
PSNR and SSIM measures and particularly the better visual quality. A new joint
NSS-base QMC method is also presented to solve the color video inpainting
problem based quaternion tensor representation. The numerical experiments on
large-scale color images and videos indicate the advantages of NSS-based QMC
over the state-of-the-art methods. | [
"cs.CV",
"cs.NA",
"math.NA",
"65F55",
"G.1.3"
] |
Sparse representation-based classifiers have shown outstanding accuracy and
robustness in image classification tasks even with the presence of intense
noise and occlusion. However, it has been discovered that the performance
degrades significantly either when test image is not aligned with the
dictionary atoms or the dictionary atoms themselves are not aligned with each
other, in which cases the sparse linear representation assumption fails. In
this paper, having both training and test images misaligned, we introduce a
novel sparse coding framework that is able to efficiently adapt the dictionary
atoms to the test image via large displacement optical flow. In the proposed
algorithm, every dictionary atom is automatically aligned with the input image
and the sparse code is then recovered using the adapted dictionary atoms. A
corresponding supervised dictionary learning algorithm is also developed for
the proposed framework. Experimental results on digit datasets recognition
verify the efficacy and robustness of the proposed algorithm. | [
"cs.CV"
] |
In dense foggy scenes, existing optical flow methods are erroneous. This is
due to the degradation caused by dense fog particles that break the optical
flow basic assumptions such as brightness and gradient constancy. To address
the problem, we introduce a semi-supervised deep learning technique that
employs real fog images without optical flow ground-truths in the training
process. Our network integrates the domain transformation and optical flow
networks in one framework. Initially, given a pair of synthetic fog images, its
corresponding clean images and optical flow ground-truths, in one training
batch we train our network in a supervised manner. Subsequently, given a pair
of real fog images and a pair of clean images that are not corresponding to
each other (unpaired), in the next training batch, we train our network in an
unsupervised manner. We then alternate the training of synthetic and real data
iteratively. We use real data without ground-truths, since to have
ground-truths in such conditions is intractable, and also to avoid the
overfitting problem of synthetic data training, where the knowledge learned on
synthetic data cannot be generalized to real data testing. Together with the
network architecture design, we propose a new training strategy that combines
supervised synthetic-data training and unsupervised real-data training.
Experimental results show that our method is effective and outperforms the
state-of-the-art methods in estimating optical flow in dense foggy scenes. | [
"cs.CV"
] |
Deep learning usually requires big data, with respect to both volume and
variety. However, most remote sensing applications only have limited training
data, of which a small subset is labeled. Herein, we review three
state-of-the-art approaches in deep learning to combat this challenge. The
first topic is transfer learning, in which some aspects of one domain, e.g.,
features, are transferred to another domain. The next is unsupervised learning,
e.g., autoencoders, which operate on unlabeled data. The last is generative
adversarial networks, which can generate realistic looking data that can fool
the likes of both a deep learning network and human. The aim of this article is
to raise awareness of this dilemma, to direct the reader to existing work and
to highlight current gaps that need solving. | [
"cs.CV",
"cs.LG",
"stat.ML"
] |
In interventional radiology, short video sequences of vein structure in
motion are captured in order to help medical personnel identify vascular issues
or plan intervention. Semantic segmentation can greatly improve the usefulness
of these videos by indicating exact position of vessels and instruments, thus
reducing the ambiguity. We propose a real-time segmentation method for these
tasks, based on U-Net network trained in a Siamese architecture from
automatically generated annotations. We make use of noisy low level binary
segmentation and optical flow to generate multi class annotations that are
successively improved in a multistage segmentation approach. We significantly
improve the performance of a state of the art U-Net at the processing speeds of
90fps. | [
"cs.CV"
] |
The main obstacle to weakly supervised semantic image segmentation is the
difficulty of obtaining pixel-level information from coarse image-level
annotations. Most methods based on image-level annotations use localization
maps obtained from the classifier, but these only focus on the small
discriminative parts of objects and do not capture precise boundaries.
FickleNet explores diverse combinations of locations on feature maps created by
generic deep neural networks. It selects hidden units randomly and then uses
them to obtain activation scores for image classification. FickleNet implicitly
learns the coherence of each location in the feature maps, resulting in a
localization map which identifies both discriminative and other parts of
objects. The ensemble effects are obtained from a single network by selecting
random hidden unit pairs, which means that a variety of localization maps are
generated from a single image. Our approach does not require any additional
training steps and only adds a simple layer to a standard convolutional neural
network; nevertheless it outperforms recent comparable techniques on the Pascal
VOC 2012 benchmark in both weakly and semi-supervised settings. | [
"cs.CV"
] |
Contrastive learning has shown superior performance in embedding global and
spatial invariant features in computer vision (e.g., image classification).
However, its overall success of embedding local and spatial variant features is
still limited, especially for semantic segmentation. In a per-pixel prediction
task, more than one label can exist in a single image for segmentation (e.g.,
an image contains both cat, dog, and grass), thereby it is difficult to define
'positive' or 'negative' pairs in a canonical contrastive learning setting. In
this paper, we propose an attention-guided supervised contrastive learning
approach to highlight a single semantic object every time as the target. With
our design, the same image can be embedded to different semantic clusters with
semantic attention (i.e., coerce semantic masks) as an additional input
channel. To achieve such attention, a novel two-stage training strategy is
presented. We evaluate the proposed method on multi-organ medical image
segmentation task, as our major task, with both in-house data and BTCV 2015
datasets. Comparing with the supervised and semi-supervised training
state-of-the-art in the backbone of ResNet-50, our proposed pipeline yields
substantial improvement of 5.53% and 6.09% in Dice score for both medical image
segmentation cohorts respectively. The performance of the proposed method on
natural images is assessed via PASCAL VOC 2012 dataset, and achieves 2.75%
substantial improvement. | [
"cs.CV",
"cs.AI",
"cs.LG"
] |
Depth from a monocular video can enable billions of devices and robots with a
single camera to see the world in 3D. In this paper, we present an approach
with a differentiable flow-to-depth layer for video depth estimation. The model
consists of a flow-to-depth layer, a camera pose refinement module, and a depth
fusion network. Given optical flow and camera pose, our flow-to-depth layer
generates depth proposals and the corresponding confidence maps by explicitly
solving an epipolar geometry optimization problem. Our flow-to-depth layer is
differentiable, and thus we can refine camera poses by maximizing the
aggregated confidence in the camera pose refinement module. Our depth fusion
network can utilize depth proposals and their confidence maps inferred from
different adjacent frames to produce the final depth map. Furthermore, the
depth fusion network can additionally take the depth proposals generated by
other methods to improve the results further. The experiments on three public
datasets show that our approach outperforms state-of-the-art depth estimation
methods, and has reasonable cross dataset generalization capability: our model
trained on KITTI still performs well on the unseen Waymo dataset. | [
"cs.CV"
] |
Most common mechanistic models are traditionally presented in mathematical
forms to explain a given physical phenomenon. Machine learning algorithms, on
the other hand, provide a mechanism to map the input data to output without
explicitly describing the underlying physical process that generated the data.
We propose a Data-based Physics Discovery (DPD) framework for automatic
discovery of governing equations from observed data. Without a prior definition
of the model structure, first a free-form of the equation is discovered, and
then calibrated and validated against the available data. In addition to the
observed data, the DPD framework can utilize available prior physical models,
and domain expert feedback. When prior models are available, the DPD framework
can discover an additive or multiplicative correction term represented
symbolically. The correction term can be a function of the existing input
variable to the prior model, or a newly introduced variable. In case a prior
model is not available, the DPD framework discovers a new data-based standalone
model governing the observations. We demonstrate the performance of the
proposed framework on a real-world application in the aerospace industry. | [
"cs.LG"
] |
Current one-stage methods for visual grounding encode the language query as
one holistic sentence embedding before fusion with visual feature. Such a
formulation does not treat each word of a query sentence on par when modeling
language to visual attention, therefore prone to neglect words which are less
important for sentence embedding but critical for visual grounding. In this
paper we propose Word2Pix: a one-stage visual grounding network based on
encoder-decoder transformer architecture that enables learning for textual to
visual feature correspondence via word to pixel attention. The embedding of
each word from the query sentence is treated alike by attending to visual
pixels individually instead of single holistic sentence embedding. In this way,
each word is given equivalent opportunity to adjust the language to vision
attention towards the referent target through multiple stacks of transformer
decoder layers. We conduct the experiments on RefCOCO, RefCOCO+ and RefCOCOg
datasets and the proposed Word2Pix outperforms existing one-stage methods by a
notable margin. The results obtained also show that Word2Pix surpasses
two-stage visual grounding models, while at the same time keeping the merits of
one-stage paradigm namely end-to-end training and real-time inference speed
intact. | [
"cs.CV",
"cs.AI",
"cs.CL"
] |
This paper proposes an automated method to obtain the extrinsic calibration
parameters between a camera and a 3D lidar with as low as 16 beams. We use a
checkerboard as a reference to obtain features of interest in both sensor
frames. The calibration board centre point and normal vector are automatically
extracted from the lidar point cloud by exploiting the geometry of the board.
The corresponding features in the camera image are obtained from the camera's
extrinsic matrix. We explain the reasons behind selecting these features, and
why they are more robust compared to other possibilities. To obtain the optimal
extrinsic parameters, we choose a genetic algorithm to address the highly
non-linear state space. The process is automated after defining the bounds of
the 3D experimental region relative to the lidar, and the true board
dimensions. In addition, the camera is assumed to be intrinsically calibrated.
Our method requires a minimum of 3 checkerboard poses, and the calibration
accuracy is demonstrated by evaluating our algorithm using real world and
simulated features. | [
"cs.CV",
"cs.RO"
] |
In many decision-making tasks, some specific actions are limited in their
frequency or total amounts, such as "fire" in the gunfight game and "buy/sell"
in the stock trading. We name such actions as "sparse action". Sparse action
often plays a crucial role in achieving good performance. However, their
Q-values, estimated by \emph{classical Bellman update}, usually suffer from a
large estimation error due to the sparsity of their samples. The \emph{greedy}
policy could be greatly misled by the biased Q-function and takes sparse action
aggressively, which leads to a huge sub-optimality. This paper constructs a
reference distribution that assigns a low probability to sparse action and
proposes a regularized objective with an explicit constraint to the reference
distribution. Furthermore, we derive a regularized Bellman operator and a
regularized optimal policy that can slow down the propagation of error and
guide the agent to take sparse action more carefully. The experiment results
demonstrate that our method achieves state-of-the-art performance on typical
sparse action tasks. | [
"cs.LG",
"cs.AI"
] |
We consider the problem of learning from sparse and underspecified rewards,
where an agent receives a complex input, such as a natural language
instruction, and needs to generate a complex response, such as an action
sequence, while only receiving binary success-failure feedback. Such
success-failure rewards are often underspecified: they do not distinguish
between purposeful and accidental success. Generalization from underspecified
rewards hinges on discounting spurious trajectories that attain accidental
success, while learning from sparse feedback requires effective exploration. We
address exploration by using a mode covering direction of KL divergence to
collect a diverse set of successful trajectories, followed by a mode seeking KL
divergence to train a robust policy. We propose Meta Reward Learning (MeRL) to
construct an auxiliary reward function that provides more refined feedback for
learning. The parameters of the auxiliary reward function are optimized with
respect to the validation performance of a trained policy. The MeRL approach
outperforms our alternative reward learning technique based on Bayesian
Optimization, and achieves the state-of-the-art on weakly-supervised semantic
parsing. It improves previous work by 1.2% and 2.4% on WikiTableQuestions and
WikiSQL datasets respectively. | [
"cs.LG",
"cs.AI",
"cs.CL",
"stat.ML"
] |
Current neural networks architectures are many times harder to train because
of the increasing size and complexity of the used datasets. Our objective is to
design more efficient training algorithms utilizing causal relationships
inferred from neural networks. The transfer entropy (TE) was initially
introduced as an information transfer measure used to quantify the statistical
coherence between events (time series). Later, it was related to causality,
even if they are not the same. There are only few papers reporting applications
of causality or TE in neural networks. Our contribution is an
information-theoretical method for analyzing information transfer between the
nodes of feedforward neural networks. The information transfer is measured by
the TE of feedback neural connections. Intuitively, TE measures the relevance
of a connection in the network and the feedback amplifies this connection. We
introduce a backpropagation type training algorithm that uses TE feedback
connections to improve its performance. | [
"cs.LG",
"cs.AI"
] |
Recent advances in self-supervised learning (SSL) have largely closed the gap
with supervised ImageNet pretraining. Despite their success these methods have
been primarily applied to unlabeled ImageNet images, and show marginal gains
when trained on larger sets of uncurated images. We hypothesize that current
SSL methods perform best on iconic images, and struggle on complex scene images
with many objects. Analyzing contrastive SSL methods shows that they have poor
visual grounding and receive poor supervisory signal when trained on scene
images. We propose Contrastive Attention-Supervised Tuning(CAST) to overcome
these limitations. CAST uses unsupervised saliency maps to intelligently sample
crops, and to provide grounding supervision via a Grad-CAM attention loss.
Experiments on COCO show that CAST significantly improves the features learned
by SSL methods on scene images, and further experiments show that CAST-trained
models are more robust to changes in backgrounds. | [
"cs.CV",
"cs.AI",
"cs.LG"
] |
Generative adversarial networks (GANs) are capable of producing high quality
image samples. However, unlike variational autoencoders (VAEs), GANs lack
encoders that provide the inverse mapping for the generators, i.e., encode
images back to the latent space. In this work, we consider adversarially
learned generative models that also have encoders. We evaluate models based on
their ability to produce high quality samples and reconstructions of real
images. Our main contributions are twofold: First, we find that the baseline
Bidirectional GAN (BiGAN) can be improved upon with the addition of an
autoencoder loss, at the expense of an extra hyper-parameter to tune. Second,
we show that comparable performance to BiGAN can be obtained by simply training
an encoder to invert the generator of a normal GAN. | [
"stat.ML",
"cs.AI",
"cs.LG"
] |
We present a novel learning framework for vehicle recognition from a single
RGB image. Unlike existing methods which only use attention mechanisms to
locate 2D discriminative information, our work learns a novel 3D perspective
feature representation of a vehicle, which is then fused with 2D appearance
feature to predict the category. The framework is composed of a global network
(GN), a 3D perspective network (3DPN), and a fusion network. The GN is used to
locate the region of interest (RoI) and generate the 2D global feature. With
the assistance of the RoI, the 3DPN estimates the 3D bounding box under the
guidance of the proposed vanishing point loss, which provides a perspective
geometry constraint. Then the proposed 3D representation is generated by
eliminating the viewpoint variance of the 3D bounding box using perspective
transformation. Finally, the 3D and 2D feature are fused to predict the
category of the vehicle. We present qualitative and quantitative results on the
vehicle classification and verification tasks in the BoxCars dataset. The
results demonstrate that, by learning such a concise 3D representation, we can
achieve superior performance to methods that only use 2D information while
retain 3D meaningful information without the challenge of requiring a 3D CAD
model. | [
"cs.CV"
] |
The artificial neural network shows powerful ability of inference, but it is
still criticized for lack of interpretability and prerequisite needs of big
dataset. This paper proposes the Rule-embedded Neural Network (ReNN) to
overcome the shortages. ReNN first makes local-based inferences to detect local
patterns, and then uses rules based on domain knowledge about the local
patterns to generate rule-modulated map. After that, ReNN makes global-based
inferences that synthesizes the local patterns and the rule-modulated map. To
solve the optimization problem caused by rules, we use a two-stage optimization
strategy to train the ReNN model. By introducing rules into ReNN, we can
strengthen traditional neural networks with long-term dependencies which are
difficult to learn with limited empirical dataset, thus improving inference
accuracy. The complexity of neural networks can be reduced since long-term
dependencies are not modeled with neural connections, and thus the amount of
data needed to optimize the neural networks can be reduced. Besides, inferences
from ReNN can be analyzed with both local patterns and rules, and thus have
better interpretability. In this paper, ReNN has been validated with a
time-series detection problem. | [
"cs.LG",
"cs.NE",
"stat.ML"
] |
Autoencoders are widely used for unsupervised learning and as a
regularization scheme in semi-supervised learning. However, theoretical
understanding of their generalization properties and of the manner in which
they can assist supervised learning has been lacking. We utilize recent
advances in the theory of deep learning generalization, together with a novel
reconstruction loss, to provide generalization bounds for autoencoders. To the
best of our knowledge, this is the first such bound. We further show that,
under appropriate assumptions, an autoencoder with good generalization
properties can improve any semi-supervised learning scheme. We support our
theoretical results with empirical demonstrations. | [
"stat.ML",
"cs.LG"
] |
Graphs can be used to represent and reason about real world systems and a
variety of metrics have been devised to quantify their global characteristics.
An important property is robustness to failures and attacks, which is relevant
for the infrastructure and communication networks that power modern society.
Prior work on making topological modifications to a graph, e.g., adding edges,
in order to increase robustness is typically based on local and spectral
properties or a shallow search since robustness is expensive to compute
directly. However, such strategies are necessarily suboptimal.
In this work, we present RNet-DQN, an approach for constructing networks that
uses Reinforcement Learning to address improving the robustness of graphs to
random and targeted removals of nodes. In particular, the approach relies on
changes in the estimated robustness as a reward signal and Graph Neural
Networks for representing states. Experiments on synthetic and real-world
graphs show that this approach can deliver performance superior to existing
methods while being much cheaper to evaluate and generalizing to out-of-sample
graphs, as well as to larger out-of-distribution graphs in some cases. The
approach is readily applicable to optimizing other global structural properties
of graphs. | [
"cs.LG",
"cs.AI",
"stat.ML"
] |
Here we demonstrate how Deep Neural Network (DNN) detections of multiple
constitutive or component objects that are part of a larger, more complex, and
encompassing feature can be spatially fused to improve the search, detection,
and retrieval (ranking) of the larger complex feature. First, scores computed
from a spatial clustering algorithm are normalized to a reference space so that
they are independent of image resolution and DNN input chip size. Then,
multi-scale DNN detections from various component objects are fused to improve
the detection and retrieval of DNN detections of a larger complex feature. We
demonstrate the utility of this approach for broad area search and detection of
Surface-to-Air Missile (SAM) sites that have a very low occurrence rate (only
16 sites) over a ~90,000 km^2 study area in SE China. The results demonstrate
that spatial fusion of multi-scale component-object DNN detections can reduce
the detection error rate of SAM Sites by $>$85% while still maintaining a 100%
recall. The novel spatial fusion approach demonstrated here can be easily
extended to a wide variety of other challenging object search and detection
problems in large-scale remote sensing image datasets. | [
"cs.CV"
] |
A common strategy to video understanding is to incorporate spatial and motion
information by fusing features derived from RGB frames and optical flow. In
this work, we introduce a new way to leverage semantic segmentation as an
intermediate representation for video understanding and use it in a way that
requires no additional labeling.
Second, we propose a general framework which learns the intermediate
representations (optical flow and semantic segmentation) jointly with the final
video understanding task and allows the adaptation of the representations to
the end goal. Despite the use of intermediate representations within the
network, during inference, no additional data beyond RGB sequences is needed,
enabling efficient recognition with a single network.
Finally, we present a way to find the optimal learning configuration by
searching the best loss weighting via evolution. We obtain more powerful visual
representations for videos which lead to performance gains over the
state-of-the-art. | [
"cs.CV"
] |
We propose a method for meta-learning reinforcement learning algorithms by
searching over the space of computational graphs which compute the loss
function for a value-based model-free RL agent to optimize. The learned
algorithms are domain-agnostic and can generalize to new environments not seen
during training. Our method can both learn from scratch and bootstrap off known
existing algorithms, like DQN, enabling interpretable modifications which
improve performance. Learning from scratch on simple classical control and
gridworld tasks, our method rediscovers the temporal-difference (TD) algorithm.
Bootstrapped from DQN, we highlight two learned algorithms which obtain good
generalization performance over other classical control tasks, gridworld type
tasks, and Atari games. The analysis of the learned algorithm behavior shows
resemblance to recently proposed RL algorithms that address overestimation in
value-based methods. | [
"cs.LG",
"cs.AI",
"cs.NE"
] |
With the success of deep neural networks, Neural Architecture Search (NAS) as
a way of automatic model design has attracted wide attention. As training every
child model from scratch is very time-consuming, recent works leverage
weight-sharing to speed up the model evaluation procedure. These approaches
greatly reduce computation by maintaining a single copy of weights on the
super-net and share the weights among every child model. However,
weight-sharing has no theoretical guarantee and its impact has not been well
studied before. In this paper, we conduct comprehensive experiments to reveal
the impact of weight-sharing: (1) The best-performing models from different
runs or even from consecutive epochs within the same run have significant
variance; (2) Even with high variance, we can extract valuable information from
training the super-net with shared weights; (3) The interference between child
models is a main factor that induces high variance; (4) Properly reducing the
degree of weight sharing could effectively reduce variance and improve
performance. | [
"cs.LG",
"cs.CV",
"stat.ML"
] |
We consider the problem of predicting edges in a graph from node attributes
in an e-commerce setting. Specifically, given nodes labelled with search query
text, we want to predict links to related queries that share products.
Experiments with a range of deep neural architectures show that simple
feedforward networks with an attention mechanism perform best for learning
embeddings. The simplicity of these models allows us to explain the performance
of attention.
We propose an analytically tractable model of query generation, AttEST, that
views both products and the query text as vectors embedded in a latent space.
We prove (and empirically validate) that the point-wise mutual information
(PMI) matrix of the AttEST query text embeddings displays a low-rank behavior
analogous to that observed in word embeddings. This low-rank property allows us
to derive a loss function that maximizes the mutual information between related
queries which is used to train an attention network to learn query embeddings.
This AttEST network beats traditional memory-based LSTM architectures by over
20% on F-1 score. We justify this out-performance by showing that the weights
from the attention mechanism correlate strongly with the weights of the best
linear unbiased estimator (BLUE) for the product vectors, and conclude that
attention plays an important role in variance reduction. | [
"cs.LG",
"stat.ML"
] |
During the last years, the emerging field of Augmented & Virtual Reality
(AR-VR) has seen tremendousgrowth. At the same time there is a trend to develop
low cost high-quality AR systems where computing poweris in demand. Feature
points are extensively used in these real-time frame-rate and 3D applications,
thereforeefficient high-speed feature detectors are necessary. Corners are such
special features and often are used as thefirst step in the marker alignment in
Augmented Reality (AR). Corners are also used in image registration
andrecognition, tracking, SLAM, robot path finding and 2D or 3D object
detection and retrieval. Therefore thereis a large number of corner detection
algorithms but most of them are too computationally intensive for use
inreal-time applications of any complexity. Many times the border of the image
is a convex polygon. For thisspecial, but quite common case, we have developed
a specific algorithm, cMinMax. The proposed algorithmis faster, approximately
by a factor of 5 compared to the widely used Harris Corner Detection algorithm.
Inaddition is highly parallelizable. The algorithm is suitable for the fast
registration of markers in augmentedreality systems and in applications where a
computationally efficient real time feature detector is necessary.The algorithm
can also be extended to N-dimensional polyhedrons. | [
"cs.CV",
"cs.GR"
] |
We introduce ThreeDWorld (TDW), a platform for interactive multi-modal
physical simulation. With TDW, users can simulate high-fidelity sensory data
and physical interactions between mobile agents and objects in a wide variety
of rich 3D environments. TDW has several unique properties: 1) realtime near
photo-realistic image rendering quality; 2) a library of objects and
environments with materials for high-quality rendering, and routines enabling
user customization of the asset library; 3) generative procedures for
efficiently building classes of new environments 4) high-fidelity audio
rendering; 5) believable and realistic physical interactions for a wide variety
of material types, including cloths, liquid, and deformable objects; 6) a range
of "avatar" types that serve as embodiments of AI agents, with the option for
user avatar customization; and 7) support for human interactions with VR
devices. TDW also provides a rich API enabling multiple agents to interact
within a simulation and return a range of sensor and physics data representing
the state of the world. We present initial experiments enabled by the platform
around emerging research directions in computer vision, machine learning, and
cognitive science, including multi-modal physical scene understanding,
multi-agent interactions, models that "learn like a child", and attention
studies in humans and neural networks. The simulation platform will be made
publicly available. | [
"cs.CV",
"cs.GR",
"cs.LG",
"cs.RO"
] |
Deep learning has made significant impacts on multi-view stereo systems.
State-of-the-art approaches typically involve building a cost volume, followed
by multiple 3D convolution operations to recover the input image's pixel-wise
depth. While such end-to-end learning of plane-sweeping stereo advances public
benchmarks' accuracy, they are typically very slow to compute. We present
MVS2D, a highly efficient multi-view stereo algorithm that seamlessly
integrates multi-view constraints into single-view networks via an attention
mechanism. Since MVS2D only builds on 2D convolutions, it is at least 4x faster
than all the notable counterparts. Moreover, our algorithm produces precise
depth estimations, achieving state-of-the-art results on challenging benchmarks
ScanNet, SUN3D, and RGBD. Even under inexact camera poses, our algorithm still
out-performs all other algorithms. Supplementary materials and code will be
available at the project page: https://zhenpeiyang.github.io/MVS2D | [
"cs.CV"
] |
Deep convolutional networks (convnets) show a remarkable ability to learn
disentangled representations. In recent years, the generalization of deep
learning to Lie groups beyond rigid motion in $\mathbb{R}^n$ has allowed to
build convnets over datasets with non-trivial symmetries, such as patterns over
the surface of a sphere. However, one limitation of this approach is the need
to explicitly define the Lie group underlying the desired invariance property
before training the convnet. Whereas rotations on the sphere have a well-known
symmetry group ($\mathrm{SO}(3)$), the same cannot be said of many real-world
factors of variability. For example, the disentanglement of pitch, intensity
dynamics, and playing technique remains a challenging task in music information
retrieval.
This article proposes a machine learning method to discover a nonlinear
transformation of the space $\mathbb{R}^n$ which maps a collection of
$n$-dimensional vectors $(\boldsymbol{x}_i)_i$ onto a collection of target
vectors $(\boldsymbol{y}_i)_i$. The key idea is to approximate every target
$\boldsymbol{y}_i$ by a matrix--vector product of the form
$\boldsymbol{\widetilde{y}}_i = \boldsymbol{\phi}(t_i) \boldsymbol{x}_i$, where
the matrix $\boldsymbol{\phi}(t_i)$ belongs to a one-parameter subgroup of
$\mathrm{GL}_n (\mathbb{R})$. Crucially, the value of the parameter $t_i \in
\mathbb{R}$ may change between data pairs $(\boldsymbol{x}_i,
\boldsymbol{y}_i)$ and does not need to be known in advance. | [
"cs.LG",
"cs.AI",
"cs.CV",
"cs.SD",
"stat.ML"
] |
A biologically plausible computational model for color representation is
introduced. We present a mechanistic hierarchical model of neurons that not
only successfully encodes local hue, but also explicitly reveals how the
contributions of each visual cortical layer participating in the process can
lead to a hue representation. Our proposed model benefits from studies on the
visual cortex and builds a network of single-opponent and hue-selective
neurons. Local hue encoding is achieved through gradually increasing
nonlinearity in terms of cone inputs to single-opponent cells. We demonstrate
that our model's single-opponent neurons have wide tuning curves, while the
hue-selective neurons in our model V4 layer exhibit narrower tunings,
resembling those in V4 of the primate visual system. Our simulation experiments
suggest that neurons in V4 or later layers have the capacity of encoding unique
hues. Moreover, with a few examples, we present the possibility of spanning the
infinite space of physical hues by combining the hue-selective neurons in our
model. | [
"cs.CV",
"I.2.10; I.4.8; I.5.4"
] |
Much attention has been devoted recently to the development of machine
learning algorithms with the goal of improving treatment policies in
healthcare. Reinforcement learning (RL) is a sub-field within machine learning
that is concerned with learning how to make sequences of decisions so as to
optimize long-term effects. Already, RL algorithms have been proposed to
identify decision-making strategies for mechanical ventilation, sepsis
management and treatment of schizophrenia. However, before implementing
treatment policies learned by black-box algorithms in high-stakes clinical
decision problems, special care must be taken in the evaluation of these
policies.
In this document, our goal is to expose some of the subtleties associated
with evaluating RL algorithms in healthcare. We aim to provide a conceptual
starting point for clinical and computational researchers to ask the right
questions when designing and evaluating algorithms for new ways of treating
patients. In the following, we describe how choices about how to summarize a
history, variance of statistical estimators, and confounders in more ad-hoc
measures can result in unreliable, even misleading estimates of the quality of
a treatment policy. We also provide suggestions for mitigating these
effects---for while there is much promise for mining observational health data
to uncover better treatment policies, evaluation must be performed
thoughtfully. | [
"cs.LG",
"stat.ML"
] |
An Axial Shifted MLP architecture (AS-MLP) is proposed in this paper.
Different from MLP-Mixer, where the global spatial feature is encoded for the
information flow through matrix transposition and one token-mixing MLP, we pay
more attention to the local features communication. By axially shifting
channels of the feature map, AS-MLP is able to obtain the information flow from
different axial directions, which captures the local dependencies. Such an
operation enables us to utilize a pure MLP architecture to achieve the same
local receptive field as CNN-like architecture. We can also design the
receptive field size and dilation of blocks of AS-MLP, etc, just like designing
those of convolution kernels. With the proposed AS-MLP architecture, our model
obtains 83.3% Top-1 accuracy with 88M parameters and 15.2 GFLOPs on the
ImageNet-1K dataset. Such a simple yet effective architecture outperforms all
MLP-based architectures and achieves competitive performance compared to the
transformer-based architectures (e.g., Swin Transformer) even with slightly
lower FLOPs. In addition, AS-MLP is also the first MLP-based architecture to be
applied to the downstream tasks (e.g., object detection and semantic
segmentation). The experimental results are also impressive. Our proposed
AS-MLP obtains 51.5 mAP on the COCO validation set and 49.5 MS mIoU on the
ADE20K dataset, which is competitive compared to the transformer-based
architectures. Code is available at https://github.com/svip-lab/AS-MLP. | [
"cs.CV"
] |
Recently, the study on object detection in aerial images has made tremendous
progress in the community of computer vision. However, most state-of-the-art
methods tend to develop elaborate attention mechanisms for the space-time
feature calibrations with high computational complexity, while surprisingly
ignoring the importance of feature calibrations in channels. In this work, we
propose a simple yet effective Calibrated-Guidance (CG) scheme to enhance
channel communications in a feature transformer fashion, which can adaptively
determine the calibration weights for each channel based on the global feature
affinity-pairs. Specifically, given a set of feature maps, CG first computes
the feature similarity between each channel and the remaining channels as the
intermediary calibration guidance. Then, re-representing each channel by
aggregating all the channels weighted together via the guidance. Our CG can be
plugged into any deep neural network, which is named as CG-Net. To demonstrate
its effectiveness and efficiency, extensive experiments are carried out on both
oriented and horizontal object detection tasks of aerial images. Results on two
challenging benchmarks (i.e., DOTA and HRSC2016) demonstrate that our CG-Net
can achieve state-of-the-art performance in accuracy with a fair computational
overhead. https://github.com/WeiZongqi/CG-Net | [
"cs.CV"
] |
Methods based on representation learning currently hold the state-of-the-art
in many natural language processing and knowledge base inference tasks. Yet, a
major challenge is how to efficiently incorporate commonsense knowledge into
such models. A recent approach regularizes relation and entity representations
by propositionalization of first-order logic rules. However,
propositionalization does not scale beyond domains with only few entities and
rules. In this paper we present a highly efficient method for incorporating
implication rules into distributed representations for automated knowledge base
construction. We map entity-tuple embeddings into an approximately Boolean
space and encourage a partial ordering over relation embeddings based on
implication rules mined from WordNet. Surprisingly, we find that the strong
restriction of the entity-tuple embedding space does not hurt the
expressiveness of the model and even acts as a regularizer that improves
generalization. By incorporating few commonsense rules, we achieve an increase
of 2 percentage points mean average precision over a matrix factorization
baseline, while observing a negligible increase in runtime. | [
"cs.LG",
"cs.AI",
"cs.CL"
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.