Update README.md
Browse files
README.md
CHANGED
@@ -27,12 +27,29 @@ size_categories:
|
|
27 |
- 1K<n<10K
|
28 |
---
|
29 |
> [!NOTE]
|
30 |
-
>
|
31 |
>
|
32 |
# The MNIST-1D Dataset
|
33 |
|
34 |
Most machine learning models get around the same ~99% test accuracy on MNIST. Our dataset, MNIST-1D, is 100x smaller (default sample size: 4000+1000; dimensionality: 40) and does a better job of separating between models with/without nonlinearity and models with/without spatial inductive biases.
|
35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
36 |
## Dataset Creation
|
37 |
|
38 |
This version of the dataset was created by using the pickle file provided by the dataset authors in the original repository: [mnist1d_data.pkl](https://github.com/greydanus/mnist1d/blob/master/mnist1d_data.pkl) and was generated like follows:
|
@@ -52,6 +69,8 @@ test = Dataset.from_dict({"x": data["x_test"], "y":data["y_test"]})
|
|
52 |
DatasetDict({"train":train, "test":test}).push_to_hub("christopher/mnist1d")
|
53 |
```
|
54 |
|
|
|
|
|
55 |
## Dataset Usage
|
56 |
|
57 |
Using the `datasets` library:
|
|
|
27 |
- 1K<n<10K
|
28 |
---
|
29 |
> [!NOTE]
|
30 |
+
> This dataset card is based on the README file of the authors' GitHub repository: https://github.com/greydanus/mnist1d
|
31 |
>
|
32 |
# The MNIST-1D Dataset
|
33 |
|
34 |
Most machine learning models get around the same ~99% test accuracy on MNIST. Our dataset, MNIST-1D, is 100x smaller (default sample size: 4000+1000; dimensionality: 40) and does a better job of separating between models with/without nonlinearity and models with/without spatial inductive biases.
|
35 |
|
36 |
+
MNIST-1D is a core teaching dataset in Simon Prince's [Understanding Deep Learning](https://udlbook.github.io/udlbook/) textbook.
|
37 |
+
|
38 |
+

|
39 |
+
|
40 |
+
|
41 |
+
## Comparing MNIST and MNIST-1D
|
42 |
+
|
43 |
+
| Dataset | Logistic Regression | MLP | CNN | GRU* | Human Expert |
|
44 |
+
|----------------------|---------------------|------|------|------|--------------|
|
45 |
+
| MNIST | 92% | 99+% | 99+% | 99+% | 99+% |
|
46 |
+
| MNIST-1D | 32% | 68% | 94% | 91% | 96% |
|
47 |
+
| MNIST-1D (shuffle**) | 32% | 68% | 56% | 57% | ~30% |
|
48 |
+
*Training the GRU takes at least 10x the walltime of the CNN.
|
49 |
+
|
50 |
+
**The term "shuffle" refers to shuffling the spatial dimension of the dataset, as in [Zhang et al. (2017)](https://arxiv.org/abs/1611.03530).
|
51 |
+
|
52 |
+
|
53 |
## Dataset Creation
|
54 |
|
55 |
This version of the dataset was created by using the pickle file provided by the dataset authors in the original repository: [mnist1d_data.pkl](https://github.com/greydanus/mnist1d/blob/master/mnist1d_data.pkl) and was generated like follows:
|
|
|
69 |
DatasetDict({"train":train, "test":test}).push_to_hub("christopher/mnist1d")
|
70 |
```
|
71 |
|
72 |
+
The origina
|
73 |
+
|
74 |
## Dataset Usage
|
75 |
|
76 |
Using the `datasets` library:
|