abstract
stringlengths 42
2.09k
|
|---|
Subaru Strategic Program with the Hyper-Suprime Cam (HSC-SSP) has proven to
be successful with its extremely-wide area coverage in past years. Taking
advantages of this feature, we report initial results from exploration and
research of expansive over- and under-dense structures at $z=$ 0.3-1 based on
the second Public Data Release where optical 5-band photometric data for $\sim$
eight million sources with $i<23$ mag are available over $\sim360$ square
degrees. We not only confirm known superclusters but also find candidates of
titanic over- and under-dense regions out to $z=1$. The mock data analysis
suggests that the density peaks would involve one or more massive dark matter
haloes ($>10^{14}$ M$_\odot$) of the redshift, and the density troughs tend to
be empty of massive haloes over $>10$ comoving Mpc. Besides, the density peaks
and troughs at $z<0.6$ are in part identified as positive and negative weak
lensing signals respectively, in mean tangential shear profiles, showing a good
agreement with those inferred from the full-sky weak lensing simulation. The
coming extensive spectroscopic surveys will be able to resolve these colossal
structures in three-dimensional space. The number density information over the
entire survey field will be available as grid-point data on the website of the
HSC-SSP data release (https://hsc.mtk.nao.ac.jp/ssp/data-release/).
|
Pre-training and fine-tuning have achieved remarkable success in many
downstream natural language processing (NLP) tasks. Recently, pre-training
methods tailored for information retrieval (IR) have also been explored, and
the latest success is the PROP method which has reached new SOTA on a variety
of ad-hoc retrieval benchmarks. The basic idea of PROP is to construct the
\textit{representative words prediction} (ROP) task for pre-training inspired
by the query likelihood model. Despite its exciting performance, the
effectiveness of PROP might be bounded by the classical unigram language model
adopted in the ROP task construction process. To tackle this problem, we
propose a bootstrapped pre-training method (namely B-PROP) based on BERT for
ad-hoc retrieval. The key idea is to use the powerful contextual language model
BERT to replace the classical unigram language model for the ROP task
construction, and re-train BERT itself towards the tailored objective for IR.
Specifically, we introduce a novel contrastive method, inspired by the
divergence-from-randomness idea, to leverage BERT's self-attention mechanism to
sample representative words from the document. By further fine-tuning on
downstream ad-hoc retrieval tasks, our method achieves significant improvements
over baselines without pre-training or with other pre-training methods, and
further pushes forward the SOTA on a variety of ad-hoc retrieval tasks.
|
By analyzing photometric and spectroscopic time series in this paper, we show
that the pulsator V764 Mon, assumed to be the brightest RR Lyrae star in the
sky, is in fact a rapidly rotating delta Scuti star with an unusually long
dominant period (P1=0.29 d). Our spectroscopy confirmed the discovery of the
Gaia satellite about the binarity of V764 Mon. In the case of HY Com, a `bona
fide' RRc star, we present its first complete radial velocity curve.
Additionally, we found that the star continues its strong phase variation
reported before.
|
Quantum computing can efficiently simulate Hamiltonian dynamics of many-body
quantum physics, a task that is generally intractable with classical computers.
The hardness lies at the ubiquitous anti-commutative relations of quantum
operators, in corresponding with the notorious negative sign problem in
classical simulation. Intuitively, Hamiltonians with more commutative terms are
also easier to simulate on a quantum computer, and anti-commutative relations
generally cause more errors, such as in the product formula method. Here, we
theoretically explore the role of anti-commutative relation in Hamiltonian
simulation. We find that, contrary to our intuition, anti-commutative relations
could also reduce the hardness of Hamiltonian simulation. Specifically,
Hamiltonians with mutually anti-commutative terms are easy to simulate, as what
happens with ones consisting of mutually commutative terms. Such a property is
further utilized to reduce the algorithmic error or the gate complexity in the
truncated Taylor series quantum algorithm for general problems. Moreover, we
propose two modified linear combinations of unitaries methods tailored for
Hamiltonians with different degrees of anti-commutation. We numerically verify
that the proposed methods exploiting anti-commutative relations could
significantly improve the simulation accuracy of electronic Hamiltonians. Our
work sheds light on the roles of commutative and anti-commutative relations in
simulating quantum systems.
|
The deep extension of the restricted Boltzmann machine (RBM), known as the
deep Boltzmann machine (DBM), is an expressive family of machine learning
models which can serve as compact representations of complex probability
distributions. However, jointly training DBMs in the unsupervised setting has
proven to be a formidable task. A recent technique we have proposed, called
mode-assisted training, has shown great success in improving the unsupervised
training of RBMs. Here, we show that the performance gains of the mode-assisted
training are even more dramatic for DBMs. In fact, DBMs jointly trained with
the mode-assisted algorithm can represent the same data set with orders of
magnitude lower number of total parameters compared to state-of-the-art
training procedures and even with respect to RBMs, provided a fan-in network
topology is also introduced. This substantial saving in number of parameters
makes this training method very appealing also for hardware implementations.
|
Probability forecasts for binary events play a central role in many
applications. Their quality is commonly assessed with proper scoring rules,
which assign forecasts a numerical score such that a correct forecast achieves
a minimal expected score. In this paper, we construct e-values for testing the
statistical significance of score differences of competing forecasts in
sequential settings. E-values have been proposed as an alternative to p-values
for hypothesis testing, and they can easily be transformed into conservative
p-values by taking the multiplicative inverse. The e-values proposed in this
article are valid in finite samples without any assumptions on the data
generating processes. They also allow optional stopping, so a forecast user may
decide to interrupt evaluation taking into account the available data at any
time and still draw statistically valid inference, which is generally not true
for classical p-value based tests. In a case study on postprocessing of
precipitation forecasts, state-of-the-art forecasts dominance tests and
e-values lead to the same conclusions.
|
We prove an upper bound on the rank of the abelianised revised fundamental
group (called "revised first Betti number") of a compact $RCD^{*}(K,N)$ space,
in the same spirit of the celebrated Gromov-Gallot upper bound on the first
Betti number for a smooth compact Riemannian manifold with Ricci curvature
bounded below. When the synthetic lower Ricci bound is close enough to
(negative) zero and the aforementioned upper bound on the revised first Betti
number is saturated (i.e. equal to the integer part of $N$, denoted by $\lfloor
N \rfloor$), then we establish a torus stability result stating that the space
is $\lfloor N \rfloor$-rectifiable as a metric measure space, and a finite
cover must be mGH-close to an $\lfloor N \rfloor$-dimensional flat torus;
moreover, in case $N$ is an integer, we prove that the space itself is
bi-H\"older homeomorphic to a flat torus. This second result extends to the
class of non-smooth $RCD^{*}(-\delta, N)$ spaces a celebrated torus stability
theorem by Colding (later refined by Cheeger-Colding).
|
Customers of machine learning systems demand accountability from the
companies employing these algorithms for various prediction tasks.
Accountability requires understanding of system limit and condition of
erroneous predictions, as customers are often interested in understanding the
incorrect predictions, and model developers are absorbed in finding methods
that can be used to get incremental improvements to an existing system.
Therefore, we propose an accountable error characterization method, AEC, to
understand when and where errors occur within the existing black-box models.
AEC, as constructed with human-understandable linguistic features, allows the
model developers to automatically identify the main sources of errors for a
given classification system. It can also be used to sample for the set of most
informative input points for a next round of training. We perform error
detection for a sentiment analysis task using AEC as a case study. Our results
on the sample sentiment task show that AEC is able to characterize erroneous
predictions into human understandable categories and also achieves promising
results on selecting erroneous samples when compared with the uncertainty-based
sampling.
|
The paper deals with the H2-norm and associated energy or power measurements
for a class of processes known as CSVIU (Control and State Variation Increase
Uncertainty). These are system models for which a stochastic process conveys
the underlying uncertainties, and are able to give rise to cautious controls.
The paper delves into the non-controlled version and fundamental system and
norms notions associated with stochastic stability and mean-square convergence.
One pillar of the study is the connection between the finiteness of one of
these norms or a limited energy measurement growth with the corresponding
stochastic stability notions. A detectability concept ties these notions, and
the analysis of linear-positive operators plays a fundamental role. The
introduction of various H2-norms and energy measurement performance criteria
allows one to span the focus from transient to long-run behavior. As the
discount parameter turns into a counter-discount, the criteria enforce stricter
requirements on the second-moment steady state errors and on the exponential
convergence rate. A tidy connection among this H2-performance measures cast
employs a unifying vanishing discount reasoning.
|
A nilmanifold is a (left) quotient of a nilpotent Lie group by a cocompact
lattice. A hypercomplex structure on a manifold is a triple of complex
structure operators satisfying the quaternionic relations. A hypercomplex
nilmanifold is a compact quotient of a nilpotent Lie group equipped with a
left-invariant hypercomplex structure. Such a manifold admits a whole
2-dimensional sphere $S^2$ of complex structures induced by quaternions. We
prove that for any hypercomplex nilmanifold $M$ and a generic complex structure
$L\in S^2$, the complex manifold $(M,L)$ has algebraic dimension 0. A stronger
result is proven when the hypercomplex nilmanifold is abelian. Consider the Lie
algebra of left-invariant vector fields of Hodge type (1,0) on the
corresponding nilpotent Lie group with respect to some complex structure $I\in
S^2$. A hypercomplex nilmanifold is called abelian when this Lie algebra is
abelian. We prove that all complex subvarieties of $(M,L)$ for generic $L\in
S^2$ on a hypercomplex abelian nilmanifold are also hypercomplex nilmanifolds.
|
Given the increasing data collection capabilities and limited computing
resources of future collider experiments, interest in using generative neural
networks for the fast simulation of collider events is growing. In our previous
study, the Bounded Information Bottleneck Autoencoder (BIB-AE) architecture for
generating photon showers in a high-granularity calorimeter showed a high
accuracy modeling of various global differential shower distributions. In this
work, we investigate how the BIB-AE encodes this physics information in its
latent space. Our understanding of this encoding allows us to propose methods
to optimize the generation performance further, for example, by altering latent
space sampling or by suggesting specific changes to hyperparameters. In
particular, we improve the modeling of the shower shape along the particle
incident axis.
|
Recent progress in nanofabrication has led to the emergence of
three-dimensional magnetic nanostructures as a vibrant field of research. This
includes the study of three-dimensional arrays of interconnected magnetic
nanowires with tunable artificial spin-ice properties. Prominent examples of
such structures are magnetic buckyball nanoarchitectures, which consist of
ferromagnetic nanowires connected at vertex positions corresponding to those of
a C60 molecule. These structures can be regarded as prototypes for the study of
the transition from two- to three-dimensional spin-ice lattices. In spite of
their significance for three-dimensional nanomagnetism, little is known about
the micromagnetic properties of buckyball nanostructures. By means of
finite-element micromagnetic simulations, we investigate the magnetization
structures and the hysteretic properties of several sub-micron-sized magnetic
buckyballs. Similar to ordinary artificial spin ice lattices, the array can be
magnetized in a variety of zero-field states with vertices exhibiting different
degrees of magnetic frustration. Remarkably, and unlike planar geometries,
magnetically frustrated states can be reversibly created and dissolved by
applying an external magnetic field. This easiness to insert and remove
defect-like magnetic charges, made possible by the angle-selectivity of the
field-induced switching of individual nanowires, demonstrates a potentially
significant advantage of three-dimensional nanomagnetism compared to planar
geometries. The control provided by the ability to switch between ice-rule
obeying and magnetically frustrated structures could be an important feature of
future applications, including magnonic devices exploiting differences in the
fundamental frequencies of these configurations.
|
This perspective article elucidates both the importance and the implications
of relativistic spacetime crystals as well as the renormalized blended
coordinates transformation. It alludes to possible applications in materials
science, condensed matter physics and quantum gravity.
|
The task of assigning semantic classes and track identities to every pixel in
a video is called video panoptic segmentation. Our work is the first that
targets this task in a real-world setting requiring dense interpretation in
both spatial and temporal domains. As the ground-truth for this task is
difficult and expensive to obtain, existing datasets are either constructed
synthetically or only sparsely annotated within short video clips. To overcome
this, we introduce a new benchmark encompassing two datasets, KITTI-STEP, and
MOTChallenge-STEP. The datasets contain long video sequences, providing
challenging examples and a test-bed for studying long-term pixel-precise
segmentation and tracking under real-world conditions. We further propose a
novel evaluation metric Segmentation and Tracking Quality (STQ) that fairly
balances semantic and tracking aspects of this task and is more appropriate for
evaluating sequences of arbitrary length. Finally, we provide several baselines
to evaluate the status of existing methods on this new challenging dataset. We
have made our datasets, metric, benchmark servers, and baselines publicly
available, and hope this will inspire future research.
|
Third-party software, or skills, are essential components in Smart Personal
Assistants (SPA). The number of skills has grown rapidly, dominated by a
changing environment that has no clear business model. Skills can access
personal information and this may pose a risk to users. However, there is
little information about how this ecosystem works, let alone the tools that can
facilitate its study. In this paper, we present the largest systematic
measurement of the Amazon Alexa skill ecosystem to date. We study developers'
practices in this ecosystem, including how they collect and justify the need
for sensitive information, by designing a methodology to identify
over-privileged skills with broken privacy policies. We collect 199,295 Alexa
skills and uncover that around 43% of the skills (and 50% of the developers)
that request these permissions follow bad privacy practices, including
(partially) broken data permissions traceability. In order to perform this kind
of analysis at scale, we present SkillVet that leverages machine learning and
natural language processing techniques, and generates high-accuracy prediction
sets. We report a number of concerning practices including how developers can
bypass Alexa's permission system through account linking and conversational
skills, and offer recommendations on how to improve transparency, privacy and
security. Resulting from the responsible disclosure we have conducted,13% of
the reported issues no longer pose a threat at submission time.
|
We introduce a new network marker for climate network analysis. It is based
upon an available special definition of local clustering coefficient for
weighted correlation networks, which was previously introduced in the
neuroscience context and aimed at compensating for uninformative correlations
caused by indirect interactions. We modify this definition further by replacing
Pearson's pairwise correlation coefficients and Pearson's three-way partial
correlation coefficients by the respective Kendall's rank correlations. This
reduces statistical sample size requirements to compute the correlations, which
translates into the possibility of using shorter time windows and hence into
shorter response time of the real-time climate network analysis. We compare
this proposed network marker to the conventional local clustering coefficient
based on unweighted networks obtained by thresholding the correlation matrix.
We show several examples where the new marker is found to be better associated
to tropical cyclones than the unweighted local clustering coefficient.
|
A previously published covariant decomposition of the Levi-Civita tensor has
demonstrated the strong mathematical parallel between gravity and $N=2$
Yang-Mills theory. We use this to argue that that the Lorentz gauge theory has
a condensate vacuum of lower energy than the perturbative vacuum. Adapting our
previously published Clairaut-based treatment of QCD, we go on to study the
implications for second quantisation.
|
Decentralized Finance (DeFi), a blockchain powered peer-to-peer financial
system, is mushrooming. One and a half years ago the total value locked in DeFi
systems was approximately $700$m USD, now, as of September 2021, it stands at
around $100$bn USD. The frenetic evolution of the ecosystem has created
challenges in understanding the basic principles of these systems and their
security risks. In this Systematization of Knowledge (SoK) we delineate the
DeFi ecosystem along the following axes: its primitives, its operational
protocol types and its security. We provide a distinction between technical
security, which has a healthy literature, and economic security, which is
largely unexplored, connecting the latter with new models and thereby
synthesizing insights from computer science, economics and finance. Finally, we
outline the open research challenges in the ecosystem across these security
types.
|
10 {\mu}m lasing is studied in a compact CO2-He cell pressurized up to 15 atm
when optically pumped by a ~50 mJ Fe:ZnSe laser tunable around 4.3 {\mu}m. The
optimal pump wavelength and partial pressure of CO2 for generating 10 {\mu}m
pulses are found to be ~4.4 {\mu}m and 0.75 atm, respectively. Without cavity
optimization, the optical-to-optical conversion efficiency reached ~10% at a
total pressure of 7 atm. The gain lifetime is measured to be ~1 {\mu}s at
pressures above 10 atm, indicating the feasibility of using high-pressure
optically pumped CO2 for the efficient amplification of picosecond 10 {\mu}m
pulses.
|
We present a unified framework for minimizing average completion time for
many seemingly disparate online scheduling problems, such as the traveling
repairperson problems (TRP), dial-a-ride problems (DARP), and scheduling on
unrelated machines.
We construct a simple algorithm that handles all these scheduling problems,
by computing and later executing auxiliary schedules, each optimizing a certain
function on already seen prefix of the input. The optimized function resembles
a prize-collecting variant of the original scheduling problem. By a careful
analysis of the interplay between these auxiliary schedules, and later
employing the resulting inequalities in a factor-revealing linear program, we
obtain improved bounds on the competitive ratio for all these scheduling
problems.
In particular, our techniques yield a $4$-competitive deterministic algorithm
for all previously studied variants of online TRP and DARP, and a
$3$-competitive one for the scheduling on unrelated machines (also with
precedence constraints). This improves over currently best ratios for these
problems that are $5.14$ and $4$, respectively. We also show how to use
randomization to further reduce the competitive ratios to $1+2/\ln 3 < 2.821$
and $1+1/\ln 2 < 2.443$, respectively. The randomized bounds also substantially
improve the current state of the art. Our upper bound for DARP contradicts the
lower bound of 3 given by Fink et al. (Inf. Process. Lett. 2009); we pinpoint a
flaw in their proof.
|
This paper introduces the Simulated Jet Engine Bracket Dataset (SimJEB): a
new, public collection of crowdsourced mechanical brackets and accompanying
structural simulations. SimJEB is applicable to a wide range of geometry
processing tasks; the complexity of the shapes in SimJEB offer a challenge to
automated geometry cleaning and meshing, while categorical labels and
structural simulations facilitate classification and regression (i.e.
engineering surrogate modeling). In contrast to existing shape collections,
SimJEB's models are all designed for the same engineering function and thus
have consistent structural loads and support conditions. On the other hand,
SimJEB models are more complex, diverse, and realistic than the synthetically
generated datasets commonly used in parametric surrogate model evaluation. The
designs in SimJEB were derived from submissions to the GrabCAD Jet Engine
Bracket Challenge: an open engineering design competition with over 700
hand-designed CAD entries from 320 designers representing 56 countries. Each
model has been cleaned, categorized, meshed, and simulated with finite element
analysis according to the original competition specifications. The result is a
collection of 381 diverse, high-quality and application-focused designs for
advancing geometric deep learning, engineering surrogate modeling, automated
cleaning and related geometry processing tasks.
|
Several governments introduced or promoted the use of contact-tracing apps
during the ongoing COVID-19 pandemic. In Germany, the related app is called
Corona-Warn-App, and by end of 2020, it had 22.8 million downloads. Contact
tracing is a promising approach for containing the spread of the novel
coronavirus. It is only effective if there is a large user base, which brings
new challenges like app users unfamiliar with using smartphones or apps. As
Corona-Warn-App is voluntary to use, reaching many users and gaining a positive
public perception is crucial for its effectiveness. Based on app reviews and
tweets, we are analyzing the public perception of Corona-Warn-App. We collected
and analyzed all 78,963 app reviews for the Android and iOS versions from
release (June 2020) to beginning of February 2021, as well as all original
tweets until February 2021 containing #CoronaWarnApp (43,082). For the reviews,
the most common words and n-grams point towards technical issues, but it
remains unclear, to what extent this is due to the app itself, the used
Exposure Notification Framework, system settings on the user's phone, or the
user's misinterpretations of app content. For Twitter data, overall, based on
tweet content, frequent hashtags, and interactions with tweets, we conclude
that the German Twitter-sphere widely reports adopting the app and promotes its
use.
|
The proximity effect from a spin-triplet $p_x$-wave superconductor to a dirty
normal-metal has been shown to result in various unusual electromagnetic
properties, reflecting a cooperative relation between topologically protected
zero-energy quasiparticles and odd-frequency Cooper pairs. However, because of
a lack of candidate materials for spin-triplet $p_x$-wave superconductors,
observing this effect has been difficult. In this paper, we demonstrate that
the anomalous proximity effect, which is essentially equivalent to that of a
spin-triplet $p_x$-wave superconductor, can occur in a semiconductor/high-$T_c$
cuprate superconductor hybrid device in which two potentials coexist: a
spin-singlet $d$-wave pair potential and a spin--orbit coupling potential
sustaining the persistent spin-helix state. As a result, we propose an
alternative and promising route to observe the anomalous proximity effect
related to the profound nature of topologically protected quasiparticles and
odd-frequency Cooper pairs.
|
The dipole anisotropy in the Cosmic Microwave Background Radiation (CMBR) has
given a peculiar velocity vector 370 km s$^{-1}$ along
$l=264^\circ,b=48^\circ$. However, some other dipoles, for instance, from the
number counts, sky brightness or redshift distributions in large samples of
distant Active Galactic Nuclei (AGNs), have yielded values of the peculiar
velocity many times larger than that from the CMBR, though surprisingly, in all
cases the directions agreed with the CMBR dipole. Here we determine our
peculiar motion from a sample of ~0.28 million AGNs, selected from the Mid
Infra Red Active Galactic Nuclei (MIRAGN) sample comprising more than a million
sources. From this, we find a peculiar velocity, which is more than four times
the CMBR value, although the direction seems to be within $\sim 2\sigma$ of the
CMBR dipole. A genuine value of the solar peculiar velocity should be the same
irrespective of the data or the technique employed to estimate it. Therefore,
such discordant dipole amplitudes, might mean that the explanation for these
dipoles, including that of the CMBR, might in fact be something else. But, the
observed fact that the direction in all cases, is the same, though obtained
from completely independent surveys using different instruments and techniques,
by different sets of people employing different computing routines, might
nonetheless indicate that these dipoles are not merely due to some systematics,
otherwise why would they all be pointing along the same direction. It might
instead suggest a preferred direction in the Universe, implying a genuine
anisotropy, which would violate the Cosmological Principle, the core of the
modern cosmology.
|
We present 3D general relativistic magnetohydrodynamic(GRMHD) simulations of
zero angular momentum accretion around a rapidly rotating black hole, modified
by the presence of initially uniform magnetic fields. We consider serveral
angles between the magnetic field direction and the black hole spin. In the
resulting flows, the midplane dynamics are governed by magnetic
reconnection-driven turbulence in a magnetically arrested (or a nearly
arrested) state. Electromagnetic jets with outflow efficiencies ~10-200% occupy
the polar regions, reaching several hundred gravitational radii before they
dissipate due to the kink instability. The jet directions fluctuate in time and
can be tilted by as much as ~30 degrees with respect to black hole spin, but
this tilt does not depend strongly on the tilt of the initial magnetic field. A
jet forms even when there is no initial net vertical magnetic flux since
turbulent, horizon-scale fluctuations can generate a net vertical field
locally. Peak jet power is obtained for an initial magnetic field tilted by
40-80 degrees with respect to the black hole spin because this maximizes the
amount of magnetic flux that can reach the black hole. These simulations may be
a reasonable model for low luminosity black hole accretion flows such as Sgr A*
or M87.
|
Large Pre-trained Language Models (PLMs) have become ubiquitous in the
development of language understanding technology and lie at the heart of many
artificial intelligence advances. While advances reported for English using
PLMs are unprecedented, reported advances using PLMs in Hebrew are few and far
between. The problem is twofold. First, Hebrew resources available for training
NLP models are not at the same order of magnitude as their English
counterparts. Second, there are no accepted tasks and benchmarks to evaluate
the progress of Hebrew PLMs on. In this work we aim to remedy both aspects.
First, we present AlephBERT, a large pre-trained language model for Modern
Hebrew, which is trained on larger vocabulary and a larger dataset than any
Hebrew PLM before. Second, using AlephBERT we present new state-of-the-art
results on multiple Hebrew tasks and benchmarks, including: Segmentation,
Part-of-Speech Tagging, full Morphological Tagging, Named-Entity Recognition
and Sentiment Analysis. We make our AlephBERT model publicly available,
providing a single point of entry for the development of Hebrew NLP
applications.
|
The goal of this minireview is restricted to describe how the
Kolmogorov-Johnson-Mehl-Avrami model has evolved from its birth up to the
present day. The model, which dates back to the late of 1930s, has the purpose
of describing the kinetics of a phase transformation. Given the nature of this
article, although there are hundreds (if not thousands) of experimental data
concerning the most disparate topics, which are interpreted on the basis of the
KJMA model, no arguments relating to these, will be touched upon. Starting from
the ingenious concept of phantom nuclei, firstly introduced by Avrami to get
the exact kinetics, we review theoretical approaches which overcome such
concept. We will show how spatial correlation among nuclei plays a fundamental
role in these modelings.
|
The layered perovskite YBaCuFeO5 (YBCFO) is considered one of the best
candidates to high-temperature chiral multiferroics with strong magnetoelectric
coupling. In RBaCuFeO5 perovskites (R: rare-earth or Y) A-site cations are
fully ordered whereas their magnetic properties strongly depend on the
preparation process. They exhibit partial cationic disorder at the B-site that
generates a magnetic spiral stabilized through directionally assisted long
range coupling between canted locally frustrated spins. Moreover the
orientation of its magnetic spiral can be critical for the magnetoelectric
response of this chiral magnetic oxide. We have synthesized and studied
YBaCuFe1-xMnxO5 samples doped with Mn, with the aim of increasing spin-orbit
coupling effects, and found that the overall Fe/Cu cation disorder at the
B-sites can be increased by doping without changing the sample preparation
process. In YBaCuFe1-xMnxO5 samples prepared under the same conditions, the T-x
magnetic phase diagram have been constructed in the range 10K-500K combining
magnetometry, X-ray and neutron powder diffraction measurements. The tilting
angles of the spins in the collinear, {\theta}col , and spiral phases,
{\theta}spiral, barely vary with temperature. In the collinear phase
{\theta}col is also independent of the Mn content. In contrast, the presence of
Mn produces a progressive reorientation of the plane of the magnetic helix in
the incommensurate phase, capable to transform the helicoidal spin ordering
into a cycloidal one, which may critically determine the ferroelectric and
magnetoelectric behavior in these compounds. Some of the observations are of
interest for engineering and developing this family of potential
high-temperature multiferroics.
|
Despite their consequential applications, metastable states of antibranes in
warped throats are not yet fully understood. In this thesis, we provide new
information on various aspects of these metastable antibranes through
applications of the blackfold effective theory for higher-dimensional black
holes. As concrete examples, we study the conjectured metastable state of
polarised anti-D3 branes at the tip of the Klebanov-Strassler (KS) throat in
type IIB supergravity and the analogous state of polarised anti-M2 branes at
the tip of the Cvetic-Gibbons-Lu-Pope (CGLP) throat in eleven-dimensional
supergravity.
For anti-D3 branes in KS throat, we provide novel evidence for the existence
of the metastable state exactly where no-go theorems are lifted. In the
extremal limit, we recover directly in supergravity the metastable states
originally discovered by Kachru, Pearson, and Verlinde (KPV). Away from
extremality, we uncover a metastable wrapped black NS5 state. We observe that
such metastability is lost when the wrapped NS5 is heated sufficiently that its
horizon geometry resembles that of a black anti-D3. We study the classical
stability of the KPV state under generic long-wavelength deformations. We
observe that, with regards to considered perturbations and regime of
parameters, the state is classically stable. A study of anti-M2 branes in CGLP
throat reveals many similarities to that of the anti-D3 branes. We recover
directly in supergravity the Klebanov-Pufu (KP) state at extremality, and our
finite temperature results fit suggestively well with known, complementary
no-go theorems. However, we discover an unexpected, exotic pattern of thermal
transitions of the KP state different from that of the KPV.
This thesis contains also a pedagogical introduction to the blackfold
formalism, focusing on aspects immediately relevant to applications to
metastable antibranes.
|
In this paper, a relay-aided two-phase transmission protocol for the smart
factory scenario is proposed. This protocol aims at enabling all robots'
ultra-reliable target number of uplink critical data transmission within a
latency constraint by jointly optimizing the relay selection, resource block
(RB) assignment, and transmit power allocation. Such protocol design is
formulated as a mixed-integer and strictly non-convex problem where
optimization variables are mutual coupling, which is definitely challenging.
Instead of conventional methods designed for solving the problem, we leverage
the properties of the relative entropy function to equivalently transform the
problem without introducing extra constraints. As the packet error probability
requirements of each robot under two possible transmission modes are coupled in
one overall reliability constraint, the big-M technique is applied to decouple
it into two corresponding reliability constraints. One is for direct
transmission mode, and the other is for cooperative transmission mode.
Moreover, both non-convex penalty (NCP) and quadratic penalty (QP) approaches
are utilized to deal with the binary indicator constraints. Based on such
penalty methods, a sequence of penalized approximated convex problems can be
iteratively solved for sub-optimal solutions. Numerical results demonstrate the
efficiency of such two penalty methods from the perspectives of sub-optimal
values of total transmit power and convergence rate. Further, the impacts of
reliability, the number and location of relays, the number of robots, the
target number of data bits on the total power consumption are analyzed.
|
We analytically study shock wave in the Josephson transmission line (JTL) in
the presence of ohmic dissipation. When ohmic resistors shunt the Josephson
junctions (JJ) or are introduced in series with the ground capacitors the shock
is broadened. When ohmic resistors are in series with the JJ, the shock remains
sharp, same as it was in the absence of dissipation. In all the cases
considered, ohmic resistors don't influence the shock propagation velocity. We
study an alternative to the shock wave - an expansion fan - in the framework of
the simple wave approximation for the dissipationless JTL and formulate the
generalization of the approximation for the JTL with ohmic dissipation.
|
The world is facing a tough situation due to the catastrophic pandemic caused
by novel coronavirus (COVID-19). The number people affected by this virus are
increasing exponentially day by day and the number has already crossed 6.4
million. As no vaccine has been discovered yet, the early detection of patients
and isolation is the only and most effective way to reduce the spread of the
virus. Detecting infected persons from chest X-Ray by using Deep Neural
Networks, can be applied as a time and laborsaving solution. In this study, we
tried to detect Covid-19 by classification of Covid-19, pneumonia and normal
chest X-Rays. We used five different Convolutional Pre-Trained Neural Network
models (VGG16, VGG19, Xception, InceptionV3 and Resnet50) and compared their
performance. VGG16 and VGG19 shows precise performance in classification. Both
models can classify between three kinds of X-Rays with an accuracy over 92%.
Another part of our study was to find the impact of weather factors
(temperature, humidity, sun hour and wind speed) on this pandemic using
Decision Tree Regressor. We found that temperature, humidity and sun-hour
jointly hold 85.88% impact on escalation of Covid-19 and 91.89% impact on death
due to Covid-19 where humidity has 8.09% impact on death. We also tried to
predict the death of an individual based on age, gender, country, and location
due to COVID-19 using the LogisticRegression, which can predict death of an
individual with a model accuracy of 94.40%.
|
Single-particle x-ray diffractive imaging (SPI) of small (bio-)nanoparticles
(NPs) requires optimized injectors to collect sufficient diffraction patterns
to reconstruct the NP structure with high resolution. Typically,
aerodynamic-lens-stack injectors are used for single NP injection. However,
current injectors were developed for larger NPs ($\gg\!100$ nm) and their
ability to generate high-density NP beams suffers with decreasing NP size.
Here, an aerodynamic-lens-stack injector with variable geometry and the
geometry-optimization procedure are presented. The optimization for 50 nm gold
NP (AuNP) injection using a numerical simulation infrastructure capable of
calculating the carrier gas flow and the particle trajectories through the
injector is introduced. The simulations are experimentally validated using
spherical AuNPs and sucrose NPs. In addition, the optimized injector is
compared to the standard-installation "Uppsala-injector" for AuNPs and results
for these heavy particles show a shift in the particle-beam focus position
rather than a change in beam size, which results in a lower gas background for
the optimized injector. Optimized aerodynamic-lens stack injectors will allow
to increase NP beam density, reduce the gas background, discover the limits of
current injectors, and contribute to structure determination of small NPs using
SPI.
|
Despite significant improvements over the last few years, cloud-based
healthcare applications continue to suffer from poor adoption due to their
limitations in meeting stringent security, privacy, and quality of service
requirements (such as low latency). The edge computing trend, along with
techniques for distributed machine learning such as federated learning, have
gained popularity as a viable solution in such settings. In this paper, we
leverage the capabilities of edge computing in medicine by analyzing and
evaluating the potential of intelligent processing of clinical visual data at
the edge allowing the remote healthcare centers, lacking advanced diagnostic
facilities, to benefit from the multi-modal data securely. To this aim, we
utilize the emerging concept of clustered federated learning (CFL) for an
automatic diagnosis of COVID-19. Such an automated system can help reduce the
burden on healthcare systems across the world that has been under a lot of
stress since the COVID-19 pandemic emerged in late 2019. We evaluate the
performance of the proposed framework under different experimental setups on
two benchmark datasets. Promising results are obtained on both datasets
resulting in comparable results against the central baseline where the
specialized models (i.e., each on a specific type of COVID-19 imagery) are
trained with central data, and improvements of 16\% and 11\% in overall
F1-Scores have been achieved over the multi-modal model trained in the
conventional Federated Learning setup on X-ray and Ultrasound datasets,
respectively. We also discuss in detail the associated challenges,
technologies, tools, and techniques available for deploying ML at the edge in
such privacy and delay-sensitive applications.
|
The goal of this paper is to propose a framework for representing and
reasoning about the rules governing a combinatorial exchange. Such a framework
is at first interest as long as we want to build up digital marketplaces based
on auction, a widely used mechanism for automated transactions. Combinatorial
exchange is the most general case of auctions, mixing the double and
combinatorial variants: agents bid to trade bundles of goods. Hence the
framework should fulfill two requirements: (i) it should enable bidders to
express their bids on combinations of goods and (ii) it should allow describing
the rules governing some market, namely the legal bids, the allocation and
payment rules. To do so, we define a logical language in the spirit of the Game
Description Language: the Combinatorial Exchange Description Language is the
first language for describing combinatorial exchange in a logical framework.
The contribution is two-fold: first, we illustrate the general dimension by
representing different kinds of protocols, and second, we show how to reason
about auction properties in this machine-processable language.
|
Active learning continues to remain significant in the industry since it is
data efficient. Not only is it cost effective on a constrained budget,
continuous refinement of the model allows for early detection and resolution of
failure scenarios during the model development stage. Identifying and fixing
failures with the model is crucial as industrial applications demand that the
underlying model performs accurately in all foreseeable use cases. One popular
state-of-the-art technique that specializes in continuously refining the model
via failure identification is Learning Loss. Although simple and elegant, this
approach is empirically motivated. Our paper develops a foundation for Learning
Loss which enables us to propose a novel modification we call LearningLoss++.
We show that gradients are crucial in interpreting how Learning Loss works,
with rigorous analysis and comparison of the gradients between Learning Loss
and LearningLoss++. We also propose a convolutional architecture that combines
features at different scales to predict the loss. We validate LearningLoss++
for regression on the task of human pose estimation (using MPII and LSP
datasets), as done in Learning Loss. We show that LearningLoss++ outperforms in
identifying scenarios where the model is likely to perform poorly, which on
model refinement translates into reliable performance in the open world.
|
We study voltage controllable superconducting state in multi-terminal bridge
composed of the dirty superconductor/pure normal metal (SN) bilayer and pure
normal metal. In the proposed system small control current $I_{ctrl}$ flows via
normal bridge, creates voltage drop $V$ and modifies distribution function of
electrons in connected SN bilayer. In case of long normal bridge the voltage
induced nonequilibrium effects could be interpreted in terms of increased local
electron temperature. In this limit we experimentally find large sensitivity of
critical curent $I_c$ of Cu/MoN/Pt-Cu bridge to $I_{ctrl}$ and relatively large
current gain which originate from steep dependence of $I_c$ on temperature and
large $I_c$ (comparable with theoretical depairing current of superconducting
bridge). In the short normal bridge deviation from equilibrium cannot be
described by simple increase of local temperature but we also theoretically
find large sensitivity of $I_c$ to control current/voltage. In this limit we
predict existence at finite $V$ of so called in-plane Fulde-Ferrell state with
spontaneous currents in SN bilayer. We argue that its appearance is connected
with voltage induced paramagnetic response in N layer.
|
A unified explicit form for difference formulas to approximate the fractional
and classical derivatives is presented. The formula gives finite difference
approximations for any classical derivatives with a desired order of accuracy
at nodal point in the computational domain. It also gives Gr\"unwald type
approximations for fractional derivatives with arbitrary order of approximation
at any point. Thus, this explicit unifies approximations of both types of
derivatives. Moreover, for classical derivatives, it provides various finite
difference formulas such as forward, backward, central, staggered, compact,
non-compact etc. Efficient computations of the coefficients of the difference
formulas are also presented that lead to automating the solution process of
differential equations with a given higher order accuracy. Some basic
applications are presented to demonstrate the usefulness of this unified
formulation.
|
Recent strategies achieved ensembling "for free" by fitting concurrently
diverse subnetworks inside a single base network. The main idea during training
is that each subnetwork learns to classify only one of the multiple inputs
simultaneously provided. However, the question of how to best mix these
multiple inputs has not been studied so far. In this paper, we introduce MixMo,
a new generalized framework for learning multi-input multi-output deep
subnetworks. Our key motivation is to replace the suboptimal summing operation
hidden in previous approaches by a more appropriate mixing mechanism. For that
purpose, we draw inspiration from successful mixed sample data augmentations.
We show that binary mixing in features - particularly with rectangular patches
from CutMix - enhances results by making subnetworks stronger and more diverse.
We improve state of the art for image classification on CIFAR-100 and Tiny
ImageNet datasets. Our easy to implement models notably outperform data
augmented deep ensembles, without the inference and memory overheads. As we
operate in features and simply better leverage the expressiveness of large
networks, we open a new line of research complementary to previous works.
|
Effects from nonstandard corrections to Newtonian gravity, at large scale,
can be investigated using the cosmological structure formation. In particular,
it is possible to show if and how a logarithmic correction (as that induced
from nonlocal gravity) modifies the clustering properties of galaxies and of
clusters of galaxies. The thermodynamics of such systems can be used to obtain
important information about the effects of such modification on clustering. We
will compare its effects with observational data and it will be demonstrated
that the observations seem to point to a characteristic scale where such a
logarithmic correction might be in play at galactic scales. However, at larger
scales such statistical inferences are much weaker, so that a fully reliable
statistical evidence for this kind of corrections cannot be stated without
further investigations and the use of more varied and precise cosmological and
astrophysical probes.
|
Recent research has shown that non-additive image steganographic frameworks
effectively improve security performance through adjusting distortion
distribution. However, as far as we know, all of the existing non-additive
proposals are based on handcrafted policies, and can only be applied to a
specific image domain, which heavily prevent non-additive steganography from
releasing its full potentiality. In this paper, we propose an automatic
non-additive steganographic distortion learning framework called MCTSteg to
remove the above restrictions. Guided by the reinforcement learning paradigm,
we combine Monte Carlo Tree Search (MCTS) and steganalyzer-based environmental
model to build MCTSteg. MCTS makes sequential decisions to adjust distortion
distribution without human intervention. Our proposed environmental model is
used to obtain feedbacks from each decision. Due to its self-learning
characteristic and domain-independent reward function, MCTSteg has become the
first reported universal non-additive steganographic framework which can work
in both spatial and JPEG domains. Extensive experimental results show that
MCTSteg can effectively withstand the detection of both hand-crafted
feature-based and deep-learning-based steganalyzers. In both spatial and JPEG
domains, the security performance of MCTSteg steadily outperforms the state of
the art by a clear margin under different scenarios.
|
State-of-the-art approaches to calculate the electron-phonon and the
phonon-electron self-energy are based on a mean-field approximation for the
interacting electronic system. This approach introduces an overscreening error
which results in an underestimation of the electron-phonon coupling strength.
We introduce a theoretical and numerical approach for the calculation of the
phonon-electron self-energy without the overscreening error. Starting from the
out-of-equilibrium Kadanoff-Baym equations for the phonon propagator, we
discuss and compare the ovescreened (i.e., symmetrically screened) and
overscreening--free (i.e., asymmetrically screened) cases. We point out that
the difficulty in treating the latter stems from the static approximation to
the dielectric function and from the need to obtain a self-energy that
preserves the elementary scattering processes. We solve both problems in the
equilibrium case by considering a manifestly symmetric form of the correct
self-energy which can be easily calculated numerically and yields an
overscreening--free coupling strength. Finally, we describe the numerical
implementation of this treatment into the first--principles Yambo code for the
calculations of phonon linewidths.
|
Ab initio molecular dynamics (AIMD) with hybrid density functionals and plane
wave basis is computationally expensive due to the high computational cost of
exact exchange energy evaluation. Recently, we proposed a strategy to combine
adaptively compressed exchange (ACE) operator formulation and multiple time
step (MTS) integration scheme to reduce the computational cost significantly
[J. Chem. Phys. 151, 151102 (2019)]. However, it was found that the
construction of the ACE operator, which has to be done at least once in every
MD time step, is computationally expensive. In the present work, systematic
improvements are introduced to further speed-up by employing localized orbitals
for the construction of the ACE operator. By this, we could achieve a
computational speed-up of an order of magnitude for a periodic system
containing 32-water molecules. Benchmark calculations were carried out to show
the accuracy and efficiency of the method in predicting the structural and
dynamical properties of bulk water. To demonstrate the applicability,
computationally intensive free energy computations at the level of hybrid
density functional theory were performed to investigate (a) methyl formate
hydrolysis reaction in neutral aqueous medium and (b) proton transfer reaction
within the active site residues of class-C $\beta$-lactamase enzyme.
|
We theoretically investigate the apparent contact angle of droplets on liquid
infused surfaces as a function of the relative size of the wetting ridge and
the deposited droplet. We provide an intuitive geometrical interpretation
whereby the variation in the apparent contact angle is due to the rotation of
the Neumann triangle. We also derive linear and quadratic corrections to the
apparent contact angle as power series expansion in terms of pressure
differences between the lubricant, droplet and gas phases. These expressions
are much simpler and more compact compared to those previously derived by
Semprebon et al. [Soft Matter, 2017, 13, 101-110].
|
Self-attention has the promise of improving computer vision systems due to
parameter-independent scaling of receptive fields and content-dependent
interactions, in contrast to parameter-dependent scaling and
content-independent interactions of convolutions. Self-attention models have
recently been shown to have encouraging improvements on accuracy-parameter
trade-offs compared to baseline convolutional models such as ResNet-50. In this
work, we aim to develop self-attention models that can outperform not just the
canonical baseline models, but even the high-performing convolutional models.
We propose two extensions to self-attention that, in conjunction with a more
efficient implementation of self-attention, improve the speed, memory usage,
and accuracy of these models. We leverage these improvements to develop a new
self-attention model family, HaloNets, which reach state-of-the-art accuracies
on the parameter-limited setting of the ImageNet classification benchmark. In
preliminary transfer learning experiments, we find that HaloNet models
outperform much larger models and have better inference performance. On harder
tasks such as object detection and instance segmentation, our simple local
self-attention and convolutional hybrids show improvements over very strong
baselines. These results mark another step in demonstrating the efficacy of
self-attention models on settings traditionally dominated by convolutional
models.
|
We introduce an additive basis of the integral cohomology ring of the
Peterson variety which reflects the geometry of certain subvarieties of the
Peterson variety. We explain the positivity of the structure constants from a
geometric viewpoint, and provide a manifestly positive combinatorial formula
for them. We also prove that our basis coincides with the additive basis
introduced by Harada-Tymoczko.
|
We study non-linear optical effects in electron systems with and without
inversion symmetry in a Fabry-Perot cavity. General photon up- and
down-conversion processes are modeled by the coupling of a noninteracting
lattice model to two modes of the quantized light field. Effective descriptions
retaining the most relevant states are devised via downfolding and a
generalized Householder transformation. These models are used to relate the
transition amplitudes for even order photon-conversion processes to the shift
vector, a topological quantity describing the difference in polarization
between the valence and conduction band in non-centrosymmetric systems. We also
demonstrate that the truncated models, despite their small Hilbert space,
capture correlation effects induced by the photons in the electronic subsystem.
|
In recent years, blockchain has grown in popularity due to its singular
attributes, enabling the development of new innovative decentralized
applications. But when companies consider leveraging blockchain for their
applications, the plethora of possible choices and the difficulty of
integrating blockchain into architectures can hinder its adoption. Our research
project aims to ease the adoption of blockchain into companies, notably with
the construction of an automated decision process to solve this issue in which
requirements are first-class citizens, a knowledge base containing
architectural patterns and blockchains refined over time, and an architecture
generator able to process outputs into architectural stubs. This paper will
also present our current progression on this decision process, by introducing
the preliminary version that is able to choose the most suitable blockchain
between multiple choices and our process-driven benchmarking tool.
|
A leader-follower framework is proposed for multi-robot navigation of large
scale teams where the leader agents corral the follower agents. A group of
leaders is modeled as a 2D deformable object where discrete masses (i.e.,
leader robots) are interconnected by springs and dampers. A time-varying domain
is defined by the positions of leaders while the external forces induce
deformations of the domain from its nominal configuration. The team of
followers is performing coverage over the time-varying domain by employing a
perspective transformation that maps between the nominal and deformed
configurations. A decentralized control strategy is proposed where a leader
only takes local sensing information and information about its neighbors
(connected by virtual springs and dampers), and a follower only needs partial
information about leaders and information about its Delaunay neighbors.
|
In the era of digitization, different actors in agriculture produce numerous
data. Such data contains already latent historical knowledge in the domain.
This knowledge enables us to precisely study natural hazards within global or
local aspects, and then improve the risk prevention tasks and augment the
yield, which helps to tackle the challenge of growing population and changing
alimentary habits. In particular, French Plants Health Bulletins (BSV, for its
name in French Bulletin de Sant{\'e} du V{\'e}g{\'e}tal) give information about
the development stages of phytosanitary risks in agricultural production.
However, they are written in natural language, thus, machines and human cannot
exploit them as efficiently as it could be. Natural language processing (NLP)
technologies aim to automatically process and analyze large amounts of natural
language data. Since the 2010s, with the increases in computational power and
parallelization, representation learning and deep learning methods became
widespread in NLP. Recent advancements Bidirectional Encoder Representations
from Transformers (BERT) inspire us to rethink of knowledge representation and
natural language understanding in plant health management domain. The goal in
this work is to propose a BERT-based approach to automatically classify the BSV
to make their data easily indexable. We sampled 200 BSV to finetune the
pretrained BERT language models and classify them as pest or/and disease and we
show preliminary results.
|
The chiral Hamiltonian for twisted graphene bilayers is written as a
$2\times2$ matrix operator by a renormalization of the Hamiltonian that takes
into account the particle-hole symmetry. This results in an effective
Hamiltonian with an average field plus and effective non-Abelian gauge
potential. The action of the proposed renormalization maps the zero-mode region
into the ground state. Modes near zero energy have an antibonding nature in a
triangular lattice. This leads to a phase-frustration effect associated with
massive degeneration, and makes flat-bands modes similar to confined modes
observed in other bipartite lattices. Suprisingly, the proposed Hamiltonian
renormalization suggests that flat-bands at magic angles are akin to
floppy-mode bands in flexible crystals or glasses, making an unexpected
connection between rigidity topological theory and magic angle twisted
two-dimensional heterostructures physics.
|
While the original goal for developing robots is replacing humans in
dangerous and tedious tasks, the final target shall be completely mimicking the
human cognitive and motor behaviour. Hence, building detailed computational
models for the human brain is one of the reasonable ways to attain this. The
cerebellum is one of the key players in our neural system to guarantee
dexterous manipulation and coordinated movements as concluded from lesions in
that region. Studies suggest that it acts as a forward model providing
anticipatory corrections for the sensory signals based on observed
discrepancies from the reference values. While most studies consider providing
the teaching signal as error in joint-space, few studies consider the error in
task-space and even fewer consider the spiking nature of the cerebellum on the
cellular-level. In this study, a detailed cellular-level forward cerebellar
model is developed, including modeling of Golgi and Basket cells which are
usually neglected in previous studies. To preserve the biological features of
the cerebellum in the developed model, a hyperparameter optimization method
tunes the network accordingly. The efficiency and biological plausibility of
the proposed cerebellar-based controller is then demonstrated under different
robotic manipulation tasks reproducing motor behaviour observed in human
reaching experiments.
|
We compute exactly the statistics of the number of records in a discrete-time
random walk model on a line where the walker stays at a given position with a
nonzero probability $0\leq p \leq 1$, while with the complementary probability
$1-p$, it jumps to a new position with a jump length drawn from a continuous
and symmetric distribution $f_0(\eta)$. We have shown that, for arbitrary $p$,
the statistics of records up to step $N$ is completely universal, i.e.,
independent of $f_0(\eta)$ for any $N$. We also compute the connected two-time
correlation function $C_p(m_1, m_2)$ of the record-breaking events at times
$m_1$ and $m_2$ and show it is also universal for all $p$. Moreover, we
demonstrate that $C_p(m_1, m_2)< C_0(m_1, m_2)$ for all $p>0$, indicating that
a nonzero $p$ induces additional anti-correlations between record events. We
further show that these anti-correlations lead to a drastic reduction in the
fluctuations of the record numbers with increasing $p$. This is manifest in the
Fano factor, i.e. the ratio of the variance and the mean of the record number,
which we compute explicitly. We also show that an interesting scaling limit
emerges when $p \to 1$, $N \to \infty$ with the product $t = (1-p)\, N$ fixed.
We compute exactly the associated universal scaling functions for the mean,
variance and the Fano factor of the number of records in this scaling limit. .
|
Large organizations such as social media companies continually release data,
for example user images. At the same time, these organizations leverage their
massive corpora of released data to train proprietary models that give them an
edge over their competitors. These two behaviors can be in conflict as an
organization wants to prevent competitors from using their own data to
replicate the performance of their proprietary models. We solve this problem by
developing a data poisoning method by which publicly released data can be
minimally modified to prevent others from train-ing models on it. Moreover, our
method can be used in an online fashion so that companies can protect their
data in real time as they release it.We demonstrate the success of our approach
onImageNet classification and on facial recognition.
|
The Born cross sections are measured for the first time for the processes
$e^+e^-\to D_s^{*+}D_{s0}^*(2317)^- +c.c.$ and $e^+e^-\to
D_s^{*+}D_{s1}(2460)^- +c.c.$ at the center-of-mass energy $\sqrt{s}=$
4.600~GeV, 4.612~GeV, 4.626~GeV, 4.640~GeV, 4.660~GeV, 4.68~GeV, and 4.700~GeV,
and for $e^+e^-\to D_s^{*+}D_{s1}(2536)^- +c.c.$ at $\sqrt{s}=$ 4.660~GeV,
4.680~GeV, and 4.700~GeV, using data samples collected with the BESIII detector
at the BEPCII collider. No structures are observed in cross-section
distributions for any of the processes.
|
The autoencoder model uses an encoder to map data samples to a lower
dimensional latent space and then a decoder to map the latent space
representations back to the data space. Implicitly, it relies on the encoder to
approximate the inverse of the decoder network, so that samples can be mapped
to and back from the latent space faithfully. This approximation may lead to
sub-optimal latent space representations. In this work, we investigate a
decoder-only method that uses gradient flow to encode data samples in the
latent space. The gradient flow is defined based on a given decoder and aims to
find the optimal latent space representation for any given sample through
optimisation, eliminating the need of an approximate inversion through an
encoder. Implementing gradient flow through ordinary differential equations
(ODE), we leverage the adjoint method to train a given decoder. We further show
empirically that the costly integrals in the adjoint method may not be entirely
necessary. Additionally, we propose a $2^{nd}$ order ODE variant to the method,
which approximates Nesterov's accelerated gradient descent, with faster
convergence per iteration. Commonly used ODE solvers can be quite sensitive to
the integration step-size depending on the stiffness of the ODE. To overcome
the sensitivity for gradient flow encoding, we use an adaptive solver that
prioritises minimising loss at each integration step. We assess the proposed
method in comparison to the autoencoding model. In our experiments, GFE showed
a much higher data-efficiency than the autoencoding model, which can be crucial
for data scarce applications.
|
New computing technologies inspired by the brain promise fundamentally
different ways to process information with extreme energy efficiency and the
ability to handle the avalanche of unstructured and noisy data that we are
generating at an ever-increasing rate. To realise this promise requires a brave
and coordinated plan to bring together disparate research communities and to
provide them with the funding, focus and support needed. We have done this in
the past with digital technologies; we are in the process of doing it with
quantum technologies; can we now do it for brain-inspired computing?
|
In this paper we characterize the value set $\Delta$ of the $R$-modules of
the form $R+zR$ for the local ring $R$ associated to a germ $\xi$ of an
irreducible plane curve singularity with one Puiseux pair. In the particular
case of the module of K\"ahler differentials attached to $\xi$, we recover some
results of Delorme. From our characterization of $\Delta$ we introduce a proper
subset of semimodules over the value semigroup of the ring $R$. Moreover, we
provide a geometric algorithm to construct all possible semimodules in this
subset for a given value semigroup.
|
Chemotaxis, the directional locomotion of cells towards a source of a
chemical gradient, is an integral part of many biological processes - for
example, bacteria motion, single-cell or multicellular organisms development,
immune response, etc. Chemotaxis directs bacteria's movement to find food
(e.g., glucose) by swimming toward the highest concentration of food molecules.
In multicellular organisms, chemotaxis is critical to early development (e.g.,
movement of sperm towards the egg during fertilization). Chemotaxis also helps
mobilize phagocytic and immune cells at sites of infection, tissue injury, and
thus facilitates immune reactions. In this paper, we study a PDE system that
describes such biological processes in one dimension, which may correspond to a
thin channel, the setting relevant in many applications: for example,
spermatozoa progression to the ovum inside a Fallopian tube or immune response
in a blood vessel.
|
This study evaluates a suspension design of a passenger car to obtain maximum
rider's comfort when the vehicle is subjected to different road profile or road
surface condition. The challenge will be on finding a balance between the
rider's comfort and vehicle handling to optimize design parameters. The study
uses a simple passive suspension system and an active suspension model
integrated with a pneumatic actuator controlled by proportional integral
derivative (PID) controller in both quarter car and full car models having a
different degree of freedoms (DOF) and increasing degrees of complexities. The
quarter car considered as a 2-DOF model, while the full car model is a 7-DOF
model. The design process set to optimise the spring stiffnesses, damping
coefficients and actuator PID controller gains. For optimisation, the research
featured genetic algorithm optimisation technique to obtain a balanced response
of the vehicle as evaluated from the displacement, velocity and acceleration of
sprung and unsprung masses along with different human comfort and vehicle
performance criteria. The results revealed that the active suspension system
with optimised spring stiffness, damping coefficients and PID gains
demonstrated the superior riding comfort and road holding compared to a passive
suspension system.
|
Prompt isolated leptons are essential in many analyses in high-energy
particle physics but are subject to fake-lepton background, i.e. objects that
mimic the lepton signature. The fake-lepton background is difficult to estimate
from simulation and is often directly determined from data. A popular method is
the matrix method, which however suffers from several limitations. This paper
recapitulates an alternative approach based on a likelihood with Poisson
constraints and reformulates the problem from a different starting point in the
framework of Bayesian statistics. The equality of both approaches is shown and
several cases are studied in which the matrix method is limited. In addition,
the fake lepton background is recalculated and compared to the estimate with
the matrix method in an example top-quark measurement.
|
Training vision-based Urban Autonomous driving models is a challenging
problem, which is highly researched in recent times. Training such models is a
data-intensive task requiring the storage and processing of vast volumes of
(possibly redundant) driving video data. In this paper, we study the problem of
developing data-efficient autonomous driving systems. In this context, we study
the problem of multi-criteria online video frame subset selection. We study
convex optimization-based solutions and show that they are unable to provide
solutions with high weightage to the loss of selected video frames. We design a
novel convex optimization-based multi-criteria online subset selection
algorithm that uses a thresholded concave function of selection variables. We
also propose and study a submodular optimization-based algorithm. Extensive
experiments using the driving simulator CARLA show that we are able to drop 80%
of the frames while succeeding to complete 100% of the episodes w.r.t. the
model trained on 100% data, in the most difficult task of taking turns. This
results in a training time of less than 30% compared to training on the whole
dataset. We also perform detailed experiments on prediction performances of
various affordances used by the Conditional Affordance Learning (CAL) model and
show that our subset selection improves performance on the crucial affordance
"Relative Angle" during turns.
|
Thinking aloud is an effective meta-cognitive strategy human reasoners apply
to solve difficult problems. We suggest to improve the reasoning ability of
pre-trained neural language models in a similar way, namely by expanding a
task's context with problem elaborations that are dynamically generated by the
language model itself. Our main result is that dynamic problem elaboration
significantly improves the zero-shot performance of GPT-2 in a deductive
reasoning and natural language inference task: While the model uses a syntactic
heuristic for predicting an answer, it is capable (to some degree) of
generating reasoned additional context which facilitates the successful
application of its heuristic. We explore different ways of generating
elaborations, including fewshot learning, and find that their relative
performance varies with the specific problem characteristics (such as problem
difficulty). Moreover, the effectiveness of an elaboration can be explained in
terms of the degree to which the elaboration semantically coheres with the
corresponding problem. In particular, elaborations that are most faithful to
the original problem description may boost accuracy by up to 24%.
|
We establish a novel approach to computing $G$-equivariant cohomology for a
finite group $G$, and demonstrate it in the case that $G = C_{p^n}$.
For any commutative ring spectrum $R$, we prove a symmetric monoidal
reconstruction theorem for genuine $G$-$R$-modules, which records them in terms
of their geometric fixedpoints as well as gluing maps involving their Tate
cohomologies. This reconstruction theorem follows from a symmetric monoidal
stratification (in the sense of \cite{AMR-strat}); here we identify the gluing
functors of this stratification in terms of Tate cohomology.
Passing from genuine $G$-spectra to genuine $G$-$\mathbb{Z}$-modules (a.k.a.
derived Mackey functors) provides a convenient intermediate category for
calculating equivariant cohomology. Indeed, as $\mathbb{Z}$-linear Tate
cohomology is far simpler than $\mathbb{S}$-linear Tate cohomology, the above
reconstruction theorem gives a particularly simple algebraic description of
genuine $G$-$\mathbb{Z}$-modules. We apply this in the case that $G = C_{p^n}$
for an odd prime $p$, computing the Picard group of genuine
$G$-$\mathbb{Z}$-modules (and therefore that of genuine $G$-spectra) as well as
the $RO(G)$-graded and Picard-graded $G$-equivariant cohomology of a point.
|
Spin defects in silicon carbide (SiC) have attracted increasing interest due
to their excellent optical and spin properties, which are useful in quantum
information processing. In this paper, we systematically investigate the
temperature dependence of the spin properties of divacancy defects in implanted
4\emph{H}-SiC. The zero-field splitting parameter $D$, the inhomogeneous
dephasing time $T_2^{*}$, the coherence time $T_2$, and the depolarization time
$T_1$ are extensively explored in a temperature range from 5 to 300 K. Two
samples implanted with different nitrogen molecule ion fluences ($\rm
{N_2}^{+}$, $1\times 10^{14}/\rm cm^{2}$ and $1\times 10^{13}/\rm cm^{2}$) are
investigated, whose spin properties are shown to have similar
temperature-dependent behaviors. Still, the sample implanted with a lower ion
fluence has longer $T_{2}$ and $T_{1}$. We provide possible theoretical
explanations for the observed temperature-dependent dynamics. Our work promotes
the understanding of the temperature dependence of spin properties in
solid-state systems, which can be helpful for constructing wide
temperature-range thermometers based on the mature semiconductor material.
|
In addition to regular Schwabe cycles (~ 11 years), solar activity also shows
longer periods of enhanced or reduced activity. Of these, reconstructions of
the Dalton Minimum provide controversial sunspot group numbers and limited
sunspot positions, partially due to limited source record accessibility. We
analysed Stephan Prantner's sunspot observations from 1804--1844, the values of
which had only been known through estimates despite their notable chronological
coverage during the Dalton Minimum. We identified his original manuscript in
Stiftsarchiv Wilten, near Innsbruck, Austria. We reviewed his biography
(1782--1873) and located his observational sites at Wilten and Waidring, which
housed the principal telescopes for his early and late observations: a 3.5-inch
astronomical telescope and a Reichenbach 4-feet achromatic erecting telescope,
respectively. We identified 215 days of datable sunspot observations, which are
twice as much data as his estimated data in the existing database (= 115 days).
Prantner counted up to 7--9 sunspot groups per day and measured sunspot
positions, which show their distributions in both solar hemispheres. These
results strikingly emphasise the difference between the Dalton Minimum and the
Maunder Minimum as well as the similarity between the Dalton Minimum and the
modern solar cycles.
|
We present algebraic and geometric classifications of the $4$-dimensional
complex nilpotent right alternative algebras. Specifically, we find that, up to
isomorphism, there are only $9$ non-isomorphic nontrivial nilpotent right
alternative algebras. The corresponding geometric variety has dimension $13$
and it is determined by the Zariski closure of $4$ rigid algebras and one
one-parametric family of algebras.
|
This paper investigates algebraic objects equipped with an operator, such as
operated monoids, operated algebras etc. Various free object functors in these
operated contexts are explicitly constructed. For operated algebras whose
operator satisfies a set $\Phi$ of relations (usually called operated
polynomial identities (aka. OPIs)), Guo defined free objects, called free
$\Phi$-algebras, via universal algebra. Free $\Phi$-algebras over algebras are
studied in details. A mild sufficient condition is found such that $\Phi$
together with a Gr\"obner-Shirshov basis of an algebra $A$ form a
Gr\"obner-Shirshov basis of the free $\Phi$-algebra over algebra $A$ in the
sense of Guo et al.. Ample examples for which this condition holds are
provided, such as all Rota-Baxter type OPIs, a class of differential type OPIs,
averaging OPIs and Reynolds OPI.
|
Neural networks trained with SGD were recently shown to rely preferentially
on linearly-predictive features and can ignore complex, equally-predictive
ones. This simplicity bias can explain their lack of robustness out of
distribution (OOD). The more complex the task to learn, the more likely it is
that statistical artifacts (i.e. selection biases, spurious correlations) are
simpler than the mechanisms to learn. We demonstrate that the simplicity bias
can be mitigated and OOD generalization improved. We train a set of similar
models to fit the data in different ways using a penalty on the alignment of
their input gradients. We show theoretically and empirically that this induces
the learning of more complex predictive patterns. OOD generalization
fundamentally requires information beyond i.i.d. examples, such as multiple
training environments, counterfactual examples, or other side information. Our
approach shows that we can defer this requirement to an independent model
selection stage. We obtain SOTA results in visual recognition on biased data
and generalization across visual domains. The method - the first to evade the
simplicity bias - highlights the need for a better understanding and control of
inductive biases in deep learning.
|
Multi-Source Domain Adaptation (MSDA) focuses on transferring the knowledge
from multiple source domains to the target domain, which is a more practical
and challenging problem compared to the conventional single-source domain
adaptation. In this problem, it is essential to utilize the labeled source data
and the unlabeled target data to approach the conditional distribution of
semantic label on target domain, which requires the joint modeling across
different domains and also an effective domain combination scheme. The
graphical structure among different domains is useful to tackle these
challenges, in which the interdependency among various instances/categories can
be effectively modeled. In this work, we propose two types of graphical
models,i.e. Conditional Random Field for MSDA (CRF-MSDA) and Markov Random
Field for MSDA (MRF-MSDA), for cross-domain joint modeling and learnable domain
combination. In a nutshell, given an observation set composed of a query sample
and the semantic prototypes i.e. representative category embeddings) on various
domains, the CRF-MSDA model seeks to learn the joint distribution of labels
conditioned on the observations. We attain this goal by constructing a
relational graph over all observations and conducting local message passing on
it. By comparison, MRF-MSDA aims to model the joint distribution of
observations over different Markov networks via an energy-based formulation,
and it can naturally perform label prediction by summing the joint likelihoods
over several specific networks. Compared to the CRF-MSDA counterpart, the
MRF-MSDA model is more expressive and possesses lower computational cost. We
evaluate these two models on four standard benchmark data sets of MSDA with
distinct domain shift and data complexity, and both models achieve superior
performance over existing methods on all benchmarks.
|
We prove that a simple knot in the lens space $L(p,q)$ fibers if and only if
its order in homology does not divide any remainder occurring in the Euclidean
algorithm applied to the pair $(p,q)$. One corollary is that if $p=m^2$ is a
perfect square, then any simple knot of order $m$ fibers, answering a question
of Cebanu. More generally, we compute the leading coefficient of the Alexander
polynomial of a simple knot, and we describe how to construct a minimum
complexity Seifert surface for one. The methods are direct, combinatorial, and
geometric.
|
In Veltman's original view, the Standard Model with a large Higgs particle
mass of about 1 TeV was the natural completion of non-renormalizable Glashow
model. This mass was thus a second threshold for weak interactions, as the W
mass was for the non-renormalizable 4-fermion V-A theory. Today, after the
observation of the narrow scalar resonance at 125 GeV, Veltman's large-mass
idea seems to be ruled out. Yet, this is not necessarily true. Depending on the
description of SSB in $\Phi^4$ theory, and by combining analytic calculations
and lattice simulations, besides the known particle at 125 GeV, a new resonance
of the Higgs field may also show up around 700 GeV. The peculiarity, though, is
that this heavier state would couple to longitudinal vector bosons with the
same typical strength of the low-mass state and thus represent a relatively
narrow resonance. In this way, such hypothetical new resonance would naturally
fit with some excess of 4-lepton events observed by ATLAS around 680 GeV.
Analogous data from CMS are needed to confirm or disprove this interpretation.
Implications of a two-mass structure for radiative corrections are also
discussed.
|
The understanding of an offense is subjective and people may have different
opinions about the offensiveness of a comment. Moreover, offenses and hate
speech may occur through sarcasm, which hides the real intention of the comment
and makes the decision of the annotators more confusing. Therefore, providing a
well-structured annotation process is crucial to a better understanding of hate
speech and offensive language phenomena, as well as supplying better
performance for machine learning classifiers. In this paper, we describe a
corpus annotation process proposed by a linguist, a hate speech specialist, and
machine learning engineers in order to support the identification of hate
speech and offensive language on social media. In addition, we provide the
first robust dataset of this kind for the Brazilian Portuguese language. The
corpus was collected from Instagram posts of political personalities and
manually annotated, being composed by 7,000 annotated documents according to
three different layers: a binary classification (offensive versus non-offensive
language), the level of offense (highly offensive, moderately offensive, and
slightly offensive messages), and the identification regarding the target of
the discriminatory content (xenophobia, racism, homophobia, sexism, religious
intolerance, partyism, apology to the dictatorship, antisemitism, and
fatphobia). Each comment was annotated by three different annotators and
achieved high inter-annotator agreement. The proposed annotation approach is
also language and domain-independent nevertheless it is currently customized
for Brazilian Portuguese.
|
We describe the addition of Lyman-alpha resonant line transfer to our dust
continuum radiation transfer code SKIRT, verifying our implementation with
published results for spherical problems and using some self-designed
three-dimensional setups. We specifically test spatial discretization through
various grid types, including hierarchical octree grids and unstructured
Voronoi tessellations. We then use a radiation transfer post-processing model
for one of the spiral galaxies produced by the Auriga cosmological zoom
simulations to investigate the effect of spatial discretization on the
synthetic observations. We find that the calculated Lyman-alpha line profiles
exhibit an extraordinarily strong dependence on the type and resolution of the
spatial grid, rendering the results untrustworthy at best. We attribute this
effect to the large gradients in the hydrogen density distribution over small
distances, which remain significantly under-resolved in the input model. We
therefore argue that further research is needed to determine the required
spatial resolution of a hydrodynamical simulation snapshot to enable meaningful
Lyman-alpha line transfer post-processing.
|
In modern data science, it is often not enough to obtain only a data-driven
model with a good prediction quality. On the contrary, it is more interesting
to understand the properties of the model, which parts could be replaced to
obtain better results. Such questions are unified under machine learning
interpretability questions, which could be considered one of the area's raising
topics. In the paper, we use multi-objective evolutionary optimization for
composite data-driven model learning to obtain the algorithm's desired
properties. It means that whereas one of the apparent objectives is precision,
the other could be chosen as the complexity of the model, robustness, and many
others. The method application is shown on examples of multi-objective learning
of composite models, differential equations, and closed-form algebraic
expressions are unified and form approach for model-agnostic learning of the
interpretable models.
|
Multi-messenger astrophysics is becoming a major avenue to explore the
Universe, with the potential to span a vast range of redshifts. The growing
synergies between different probes is opening new frontiers, which promise
profound insights into several aspects of fundamental physics and cosmology. In
this context, THESEUS will play a central role during the 2030s in detecting
and localizing the electromagnetic counterparts of gravitational wave and
neutrino sources that the unprecedented sensitivity of next generation
detectors will discover at much higher rates than the present. Here, we review
the most important target signals from multi-messenger sources that THESEUS
will be able to detect and characterize, discussing detection rate expectations
and scientific impact.
|
We adopt a geometric perspective on Fock space to provide two complementary
insights into the eigenstates in many-body-localized fermionic systems. On the
one hand, individual many-bodylocalized eigenstates are well approximated by a
Slater determinant of single-particle orbitals. On the other hand, the orbitals
of different eigenstates in a given system display a varying, and generally
imperfect, degree of compatibility, as we quantify by a measure based on the
projectors onto the corresponding single-particle subspaces. We study this
incompatibility between states of fixed and differing particle number, as well
as inside and outside the many-body-localized regime. This gives detailed
insights into the emergence and strongly correlated nature of
quasiparticle-like excitations in many-body localized systems, revealing
intricate correlations between states of different particle number down to the
level of individual realizations.
|
In applications such as participatory sensing and crowd sensing,
self-interested agents exert costly effort towards achieving an objective for
the system operator. We study such a setup where a principal incentivizes
multiple agents of different types who can collude with each other to derive
rent. The principal cannot observe the efforts exerted directly, but only the
outcome of the task, which is a noisy function of the effort. The type of each
agent influences the effort cost and task output. For a duopoly in which agents
are coupled in their payments, we show that if the principal and the agents
interact finitely many times, the agents can derive rent by colluding even if
the principal knows the types of the agents. However, if the principal and the
agents interact infinitely often, the principal can disincentivize agent
collusion through a suitable data-driven contract.
|
We establish an exact formulation for wave scattering of a massless field
with spin and charge by a Kerr-Newman-de Sitter black hole. Our formulation is
based on the exact solution of the Teukolsky equation in terms of the local
Heun function, and does not require any approximation. It serves as simple
exact formulae with arbitrary high precision, which realize fast calculation
without restrictions on model parameters. We highlight several applications
including quasinormal modes, cross section, reflection/absorption rate, and
Green function.
|
The Sloan Digital Sky Survey provides colors for more than 100 000 moving
objects, among which around 10 000 have albedos determined. Here we combined
colors and albedo in order to perform a cluster analysis on the small bodies
population, and identify a C-cluster, a group of asteroid related to C-type as
defined in earlier work. Members of this C-cluster are in fair agreement with
the color boundaries of B and C-type defined in DeMeo and Carry (2013). We then
compare colors of C-cluster asteroids to those of carbonaceous chondrites
powders, while taking into account the effect of phase angle. We show that only
CM chondrites have colors in the range of C-cluster asteroids, CO, CR and CV
chondrites being significantly redder. Also, CM chondrites powders are on
average slightly redder than the average C-cluster. The colors of C-cluster
members are further investigated by looking at color variations as a function
of asteroid diameter. We observe that the visible slope becomes bluer with
decreasing asteroids diameter, and a transition seems to be present around 20
km. We discuss the origin of this variation and, if not related to a bias in
the dataset - analysis, we conclude that it is related to the surface texture
of the objects, smaller objects being covered by rocks, while larger objects
are covered by a particulate surface. The blueing is interpreted by an
increased contribution of the first reflection in the case of rock-dominated
surfaces, which can scatter light in a Rayleigh-like manner. We do not have
unambiguous evidence of space weathering within the C-cluster based on this
analysis, however the generally bluer nature of C-cluster objects compared to
CM chondrites could be to some extent related to space weathering.
|
An intersecting $r$-uniform straight line system is an intersecting linear
system whose lines consist of $r$ points on straight line segment of
$\mathbb{R}^2$ and any two lines share a point. Recently, the author [A.
V\'azquez-\'Avila, \emph{On intersecting straight line systems}, J. Discret.
Math. Sci. Cryptogr. Accepted] proved that any intersecting $r$-uniform
straight line system $(P,\mathcal{L})$ has transversal number at most
$\nu_2-1$, with $r\geq\nu_2$, where $\nu_2$ is the maximum cardinality of a
subset of lines $R\subseteq\mathcal{L}$ such that every triplet of different
elements of $R$ does not have a common point.
In this paper, we improve such upper bound if the intersecting $r$-uniform
straight line system satisfies $r=\nu_2$. This result has immediate
consequences for some questions given by Oliveros et al. [D. Oliveros, C.
O'Neill and S. Zerbib, \emph{The geometry and combinatorics of discrete line
segment hypergraphs}, Discrete Math. {\bf 343} (2020), no. 6, 111825].
|
The effect of polyvalent molecular cations, such as spermine, on the
condensation of DNA into very well-defined toroidal shapes have been well
studied and understood. However, a great effort has been made trying to obtain
similar condensed structures from either ssRNA or dsRNA, which the latter
carries similar negative charge density as dsDNA, although it adopts a
different helical form. But the analogous condensation of RNA molecules into
well-defined structures has so far been elusive. In this work, we show that
ssRNA molecules can easily be condensed into nanoring structures on a mice
surface, where each nanoring structure is formed mostly by a single RNA
molecule. The condensation occurs in a concentration range of different atomic
cations, from monovalent to trivalent. The structures of the RNA nanorings on
mica surfaces were oberved by atomic force microscopy (AFM). The samples were
observed in tapping mode and were prepared by drop evaporation of a solution of
RNA in the presence of one type of the different cations used. As far as we
know, this is the first time that nanorings or any other well-defined condensed
RNA structures have been reported. The RNA nanorings formation can be
understood by an energy competition between the hydrogen bonding forming
hairpin stems, weakened by the salts, and hairpin loops. This results may have
an important biological relevance, since it has been proposed that RNA is the
oldest genome coding molecule and the formation of these structures could have
given it stability against degradation in primeval times. Even more, the
nanoring structures could have the potential to be used as biosensors and
functionalized nanodevices.
|
We explore the feasibility of combining Graph Neural Network-based policy
architectures with Deep Reinforcement Learning as an approach to problems in
systems. This fits particularly well with operations on networks, which
naturally take the form of graphs. As a case study, we take the idea of
data-driven routing in intradomain traffic engineering, whereby the routing of
data in a network can be managed taking into account the data itself. The
particular subproblem which we examine is minimising link congestion in
networks using knowledge of historic traffic flows. We show through experiments
that an approach using Graph Neural Networks (GNNs) performs at least as well
as previous work using Multilayer Perceptron architectures. GNNs have the added
benefit that they allow for the generalisation of trained agents to different
network topologies with no extra work. Furthermore, we believe that this
technique is applicable to a far wider selection of problems in systems
research.
|
We seek to design experimentally feasible broadband, temporally multiplexed
optical quantum memory with near-term applications to telecom bands.
Specifically, we devise dispersion compensation for an impedance-matched
narrow-band quantum memory by exploiting Raman processes over two three-level
atomic subensembles, one for memory and the other for dispersion compensation.
Dispersion compensation provides impedance matching over more than a full
cavity linewidth. Combined with one second spin-coherence lifetime the memory
could be capable of power efficiency exceeding 90% leading to 106 modes for
temporal multiplexing. Our design could lead to significant multiplexing
enhancement for quantum repeaters to be used for telecom quantum networks.
|
The flux of cosmic rays (CRs) in the heliosphere is subjected to remarkable
time variations caused by the 11-year cycle of solar activity. To help the
study of this effect, we have developed a web application (Heliophysics Virtual
Observatory) that collects real-time data on solar activity, interplanetary
plasma, and charged radiation from several space missions or observatories. As
we will show, our application can be used to visualize, manipulate, and
download updated data on sunspots, heliospheric magnetic fields, solar wind,
and neutron monitors counting rates. Data and calculations are automatically
updated on daily basis. A nowcasting for the energy spectrum of CR protons
near-Earth is also provided using calculations and real-time neutron monitor
data as input.
|
Flickering is a universal phenomenon in accreting astronomical systems which
still defies detailed physical understanding. It is particularly evident in
cataclysmic variables (CVs). Attempting to define boundary conditions for
models, the strength of the flickering is measured in several thousand light
curves of more than 100 CVs. The flickering amplitude is parameterized by the
FWHM of a Gaussian fit to the magnitude distribution of data points in a light
curve. This quantity requires several corrections before a comparison between
different sources can be made. While no correlations of the flickering strength
with simple parameters such as component masses, orbital inclination or period
were detected, a dependence on the absolute magnitude of the primary component
and on the CV subtype is found. In particular, flickering in VY Scl tpye
novalike variables is systematically stronger than in UX UMa type novalikes.
The broadband spectrum of the flickering light source can be fit by simple
models but shows excess in the $U$ band. When the data permitted to investigate
the flickering strength as a function of orbital phase in eclipsing CVs, such a
dependence was found, but it is different for different systems. Surprisingly,
there are also indications for variations of the flickering strength with the
superhump phase in novalike variables with permanent superhumps. In dwarf
novae, the flickering amplitude is high during quiescence, drops quickly at an
intermediate magnitude when the system enters into (or returns from) an
outburst and, on average, remains constant above a given brightness threshold.
|
Training a multi-speaker Text-to-Speech (TTS) model from scratch is
computationally expensive and adding new speakers to the dataset requires the
model to be re-trained. The naive solution of sequential fine-tuning of a model
for new speakers can cause the model to have poor performance on older
speakers. This phenomenon is known as catastrophic forgetting. In this paper,
we look at TTS modeling from a continual learning perspective where the goal is
to add new speakers without forgetting previous speakers. Therefore, we first
propose an experimental setup and show that serial fine-tuning for new speakers
can result in the forgetting of the previous speakers. Then we exploit two
well-known techniques for continual learning namely experience replay and
weight regularization and we reveal how one can mitigate the effect of
degradation in speech synthesis diversity in sequential training of new
speakers using these methods. Finally, we present a simple extension to improve
the results in extreme setups.
|
The difficulty of optimal control problems has classically been characterized
in terms of system properties such as minimum eigenvalues of
controllability/observability gramians. We revisit these characterizations in
the context of the increasing popularity of data-driven techniques like
reinforcement learning (RL), and in control settings where input observations
are high-dimensional images and transition dynamics are unknown. Specifically,
we ask: to what extent are quantifiable control and perceptual difficulty
metrics of a task predictive of the performance and sample complexity of
data-driven controllers? We modulate two different types of partial
observability in a cartpole "stick-balancing" problem -- (i) the height of one
visible fixation point on the cartpole, which can be used to tune fundamental
limits of performance achievable by any controller, and by (ii) the level of
perception noise in the fixation point position inferred from depth or RGB
images of the cartpole. In these settings, we empirically study two popular
families of controllers: RL and system identification-based $H_\infty$ control,
using visually estimated system state. Our results show that the fundamental
limits of robust control have corresponding implications for the
sample-efficiency and performance of learned perception-based controllers.
Visit our project website https://jxu.ai/rl-vs-control-web for more
information.
|
We introduce and apply a new approach to probe the response of galactic
stellar haloes to the interplay between cosmological merger histories and
galaxy formation physics. We perform dark-matter-only, zoomed simulations of
two Milky Way-mass hosts and make targeted, controlled changes to their
cosmological histories using the genetic modification technique. Populating
each history's stellar halo with a semi-empirical, particle-tagging approach
then enables a controlled study, with all instances converging to the same
large-scale structure, dynamical and stellar mass at $z=0$ as their reference.
These related merger scenarios alone generate an extended spread in stellar
halo mass fractions (1.5 dex) comparable to the observed population. Largest
scatter is achieved by growing late ($z\leq1$) major mergers that spread out
existing stars to create massive, in-situ dominated stellar haloes. Increasing
a last major merger at $z\sim2$ brings more accreted stars into the inner
regions, resulting in smaller scatter in the outskirts which are predominantly
built by subsequent minor events. Exploiting the flexibility of our
semi-empirical approach, we show that the diversity of stellar halo masses
across scenarios is reduced by allowing shallower slopes in the stellar
mass--halo mass relation for dwarf galaxies, while it remains conserved when
central stars are born with hotter kinematics across cosmic time. The
merger-dependent diversity of stellar haloes thus responds distinctly to
assumptions in modelling the central and dwarf galaxies respectively, opening
exciting prospects to constrain star formation and feedback at different
galactic mass-scales with the coming generation of deep, photometric
observatories.
|
This paper presents a method for riggable 3D face reconstruction from
monocular images, which jointly estimates a personalized face rig and per-image
parameters including expressions, poses, and illuminations. To achieve this
goal, we design an end-to-end trainable network embedded with a differentiable
in-network optimization. The network first parameterizes the face rig as a
compact latent code with a neural decoder, and then estimates the latent code
as well as per-image parameters via a learnable optimization. By estimating a
personalized face rig, our method goes beyond static reconstructions and
enables downstream applications such as video retargeting. In-network
optimization explicitly enforces constraints derived from the first principles,
thus introduces additional priors than regression-based methods. Finally,
data-driven priors from deep learning are utilized to constrain the ill-posed
monocular setting and ease the optimization difficulty. Experiments demonstrate
that our method achieves SOTA reconstruction accuracy, reasonable robustness
and generalization ability, and supports standard face rig applications.
|
Object recognition has made great advances in the last decade, but
predominately still relies on many high-quality training examples per object
category. In contrast, learning new objects from only a few examples could
enable many impactful applications from robotics to user personalization. Most
few-shot learning research, however, has been driven by benchmark datasets that
lack the high variation that these applications will face when deployed in the
real-world. To close this gap, we present the ORBIT dataset and benchmark,
grounded in the real-world application of teachable object recognizers for
people who are blind/low-vision. The dataset contains 3,822 videos of 486
objects recorded by people who are blind/low-vision on their mobile phones. The
benchmark reflects a realistic, highly challenging recognition problem,
providing a rich playground to drive research in robustness to few-shot,
high-variation conditions. We set the benchmark's first state-of-the-art and
show there is massive scope for further innovation, holding the potential to
impact a broad range of real-world vision applications including tools for the
blind/low-vision community. We release the dataset at
https://doi.org/10.25383/city.14294597 and benchmark code at
https://github.com/microsoft/ORBIT-Dataset.
|
The recent inflow of empirical data about the collective behaviour of
strongly correlated biological systems has brought field theory and the
renormalization group into the biophysical arena. Experiments on bird flocks
and insect swarms show that social forces act on the particles' velocity
through the generator of its rotations, namely the spin, indicating that
mode-coupling field theories are necessary to reproduce the correct dynamical
behaviour. Unfortunately, a theory for three coupled fields - density, velocity
and spin - has a prohibitive degree of intricacy. A simplifying path consists
in getting rid of density fluctuations by studying incompressible systems. This
requires imposing a solenoidal constraint on the primary field, an unsolved
problem even for equilibrium mode-coupling theories. Here, we perform an
equilibrium dynamic renormalization group analysis of a mode-coupling field
theory subject to a solenoidal constraint; using the classification of Halperin
and Hohenberg, we can dub this case as a solenoidal Model G. We demonstrate
that the constraint produces a new vertex that mixes static and dynamical
coupling constants, and that this vertex is essential to grant the closure of
the renormalization group structure and the consistency of dynamics with
statics. Interestingly, although the solenoidal constraint leads to a
modification of the static universality class, we find that it does not change
the dynamical universality class, a result that seems to represent an exception
to the general rule that dynamical universality classes are narrower than
static ones. Our results constitute a solid stepping stone in the admittedly
large chasm towards developing an off-equilibrium mode-coupling theory of
biological groups.
|
Besides magnetic and charge order, regular arrangements of orbital occupation
constitute a fundamental order parameter of condensed matter physics. Even
though orbital order is difficult to identify directly in experiments, its
presence was firmly established in a number of strongly correlated,
three-dimensional Mott insulators. Here, reporting resonant X-ray scattering
experiments on the layered Van der Waals compound $1T$-TiSe$_2$, we establish
the emergence of orbital order in a weakly correlated, quasi-two-dimensional
material. Our experimental scattering results are consistent with
first-principles calculations that bring to the fore a generic mechanism of
close interplay between charge redistribution, lattice displacements, and
orbital order. It demonstrates the essential role that orbital degrees of
freedom play in TiSe$_2$, and their importance throughout the family of
correlated Van der Waals materials.
|
Existing radar sensors can be classified into automotive and scanning radars.
While most radar odometry (RO) methods are only designed for a specific type of
radar, our RO method adapts to both scanning and automotive radars. Our RO is
simple yet effective, where the pipeline consists of thresholding,
probabilistic submap building, and an NDT-based radar scan matching. The
proposed RO has been tested on two public radar datasets: the Oxford Radar
RobotCar dataset and the nuScenes dataset, which provide scanning and
automotive radar data respectively. The results show that our approach
surpasses state-of-the-art RO using either automotive or scanning radar by
reducing translational error by 51% and 30%, respectively, and rotational error
by 17% and 29%, respectively. Besides, we show that our RO achieves
centimeter-level accuracy as lidar odometry, and automotive and scanning RO
have similar accuracy.
|
Periodically-driven systems are ubiquitous in science and technology. In
quantum dynamics, even a small number of periodically-driven spins leads to
complicated dynamics. Hence, it is of interest to understand what constraints
such dynamics must satisfy. We derive a set of constraints for each number of
cycles. For pure initial states, the observable being constrained is the
recurrence probability. We use our constraints for detecting undesired coupling
to unaccounted environments and drifts in the driving parameters. To illustrate
the relevance of these results for modern quantum systems we demonstrate our
findings experimentally on a trapped-ion quantum computer, and on various IBM
quantum computers. Specifically, we provide two experimental examples where
these constraints surpass fundamental bounds associated with known one-cycle
constraints. This scheme can potentially be used to detect the effect of the
environment in quantum circuits that cannot be classically simulated. Finally,
we show that, in practice, testing an $n$-cycle constraint requires executing
only $O(\sqrt{n})$ cycles, which makes the evaluation of constraints associated
with hundreds of cycles realistic.
|
We study one-loop bulk entanglement entropy in even spacetime dimensions
using the heat kernel method, which captures the universal piece of
entanglement entropy, a logarithmically divergent term in even dimensions. In
four dimensions, we perform explicit calculations for various shapes of
boundary subregions. In particular, for a cusp subregion with an arbitrary
opening angle, we find that the bulk entanglement entropy always encodes the
same universal information about the boundary theories as the leading
entanglement entropy in the large N limit, up to a fixed proportional constant.
By smoothly deforming a circle in the boundary, we find that to leading order
of the deformations, the bulk entanglement entropy shares the same shape
dependence as the leading entanglement entropy and hence the same physical
information can be extracted from both cases. This establishes an interesting
local/nonlocal duality for holographic $\mm{CFT}_3$. However, the result does
not hold for higher dimensional holographic theories.
|
Using the nonabilinization procedure, we find an integrable matrix version of
the Euler top on $\mathfrak{so}_3$
|
Linear-response time-dependent density functional theory (LR-TDDFT) for core
level spectroscopy using standard local functionals suffers from
self-interaction error and a lack of orbital relaxation upon creation of the
core hole. As a result, LR-TDDFT calculated X-ray absorption near edge
structure (XANES) spectra need to be shifted along the energy axis to match
experimental data. We propose a correction scheme based on many body
perturbation theory to calculate the shift from first principles. The
ionization potential of the core donor state is first computed and then
substituted for the corresponding Kohn--Sham orbital energy, thus emulating
Koopmans' condition. Both self-interaction error and orbital relaxation are
taken into account. The method exploits the localized nature of core states for
efficiency and integrates seamlessly in our previous implementation of core
level LR-TDDFT, yielding corrected spectra in a single calculation. We
benchmark the correction scheme on molecules at the K- and L-edges as well as
for core binding energies and report accuracies comparable to higher order
methods. We also demonstrate applicability in large and extended systems and
discuss efficient approximations.
|
A giant impact origin for the Moon is generally accepted, but many aspects of
lunar formation remain poorly understood and debated. \'Cuk et al. (2016)
proposed that an impact that left the Earth-Moon system with high obliquity and
angular momentum could explain the Moon's orbital inclination and isotopic
similarity to Earth. In this scenario, instability during the Laplace Plane
transition, when the Moon's orbit transitions from the gravitational influence
of Earth's figure to that of the Sun, would both lower the system's angular
momentum to its present-day value and generate the Moon's orbital inclination.
Recently, Tian and Wisdom (2020) discovered new dynamical constraints on the
Laplace Plane transition and concluded that the Earth-Moon system could not
have evolved from an initial state with high obliquity. Here we demonstrate
that the Earth-Moon system with an initially high obliquity can evolve into the
present state, and we identify a spin-orbit secular resonance as a key
dynamical mechanism in the later stages of the Laplace Plane transition. Some
of the simulations by Tian and Wisdom (2020) did not encounter this late
secular resonance, as their model suppressed obliquity tides and the resulting
inclination damping. Our results demonstrate that a giant impact that left
Earth with high angular momentum and high obliquity ($\theta > 61^{\circ}$) is
a promising scenario for explaining many properties of the Earth-Moon system,
including its angular momentum and obliquity, the geochemistry of Earth and the
Moon, and the lunar inclination.
|
Transmission electron microscopes use electrons with wavelengths of a few
picometers, potentially capable of imaging individual atoms in solids at a
resolution ultimately set by the intrinsic size of an atom. Unfortunately, due
to imperfections in the imaging lenses and multiple scattering of electrons in
the sample, the image resolution reached is 3 to 10 times worse. Here, by
inversely solving the multiple scattering problem and overcoming the
aberrations of the electron probe using electron ptychography to recover a
linear phase response in thick samples, we demonstrate an instrumental blurring
of under 20 picometers. The widths of atomic columns in the measured
electrostatic potential are now no longer limited by the imaging system, but
instead by the thermal fluctuations of the atoms. We also demonstrate that
electron ptychography can potentially reach a sub-nanometer depth resolution
and locate embedded atomic dopants in all three dimensions with only a single
projection measurement.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.