Understanding deep learning requires rethinking generalizationUnderstanding deep learning requires rethinking generalizationChiyuan Zhang and Samy Bengio and Moritz Hardt and Benjamin Recht and Oriol Vinyals2016
Paper summarymarekThe authors investigate the generalisation properties of several well-known image recognition networks.
https://i.imgur.com/km0mrVs.png
They show that these networks are able to overfit to the training set with 100% accuracy even if the labels on the images are random, or if the pixels are randomly generated. Regularisation, such as weight decay and dropout, doesn’t stop overfitting as much as expected, still resulting in ~90% accuracy on random training data. They then argue that these models likely make use of massive memorization, in combination with learning low-complexity patterns, in order to perform well on these tasks.
First published: 2016/11/10 (2 years ago) Abstract: Despite their massive size, successful deep artificial neural networks can
exhibit a remarkably small difference between training and test performance.
Conventional wisdom attributes small generalization error either to properties
of the model family, or to the regularization techniques used during training.
Through extensive systematic experiments, we show how these traditional
approaches fail to explain why large neural networks generalize well in
practice. Specifically, our experiments establish that state-of-the-art
convolutional networks for image classification trained with stochastic
gradient methods easily fit a random labeling of the training data. This
phenomenon is qualitatively unaffected by explicit regularization, and occurs
even if we replace the true images by completely unstructured random noise. We
corroborate these experimental findings with a theoretical construction showing
that simple depth two neural networks already have perfect finite sample
expressivity as soon as the number of parameters exceeds the number of data
points as it usually does in practice.
We interpret our experimental findings by comparison with traditional models.
This paper deals with the question what / how exactly CNNs learn, considering the fact that they usually have more trainable parameters than data points on which they are trained.
When the authors write "deep neural networks", they are talking about Inception V3, AlexNet and MLPs.
## Key contributions
* Deep neural networks easily fit random labels (achieving a training error of 0 and a test error which is just randomly guessing labels as expected). $\Rightarrow$Those architectures can simply brute-force memorize the training data.
* Deep neural networks fit random images (e.g. Gaussian noise) with 0 training error. The authors conclude that VC-dimension / Rademacher complexity, and uniform stability are bad explanations for generalization capabilities of neural networks
* The authors give a construction for a 2-layer network with $p = 2n+d$ parameters - where $n$ is the number of samples and $d$ is the dimension of each sample - which can easily fit any labeling. (Finite sample expressivity). See section 4.
## What I learned
* Any measure $m$ of the generalization capability of classifiers $H$ should take the percentage of corrupted labels ($p_c \in [0, 1]$, where $p_c =0$ is a perfect labeling and $p_c=1$ is totally random) into account: If $p_c = 1$, then $m()$ should be 0, too, as it is impossible to learn something meaningful with totally random labels.
* We seem to have built models which work well on image data in general, but not "natural" / meaningful images as we thought.
## Funny
> deep neural nets remain mysterious for many reasons
> Note that this is not exactly simple as the kernel matrix requires 30GB to store in memory. Nonetheless, this system can be solved in under 3 minutes in on a commodity workstation with 24 cores and 256 GB of RAM with a conventional LAPACK call.
## See also
* [Deep Nets Don't Learn Via Memorization](https://openreview.net/pdf?id=rJv6ZgHYg)
_Objective:_ Theoretical study of Deep Neural Network, their expressivity and regularizations.
## Results:
The key findings of the article are:
### A. Deep neural networks easily fit random labels.
Both when randomizing labels, replacing images with raw noise or all situations in-between.
1. The effective capacity of neural networks is sufficient for memorizing the entire data set.
2. Even optimization on random labels remains easy. In fact, training time increases only by a small constant factor compared with training on the true labels.
3. Randomizing labels is solely a data transformation, leaving all other properties of the learning problem unchanged.
### B. Explicit regularization may improve generalization performance, but is neither necessary nor by itself sufficient for controlling generalization error.
By explicit regularization they mean batch normalisation, weight decay, dropout, data augmentation, etc.
### C. Generically large neural networks can express any labeling of the training data.
More formally, a very simple two-layer ReLU network with `p = 2n + d` parameters can express any labeling of any sample of size `n` in `d` dimensions.
### D. The optimization algorithm itself is implicitly regularizing the solution.
SGD acts as an implicit regularizer and properties are inherited by models that were trained using SGD.
The authors investigate the generalisation properties of several well-known image recognition networks.
https://i.imgur.com/km0mrVs.png
They show that these networks are able to overfit to the training set with 100% accuracy even if the labels on the images are random, or if the pixels are randomly generated. Regularisation, such as weight decay and dropout, doesn’t stop overfitting as much as expected, still resulting in ~90% accuracy on random training data. They then argue that these models likely make use of massive memorization, in combination with learning low-complexity patterns, in order to perform well on these tasks.