Deep Learning without Shortcuts: Shaping the Kernel with Tailored Rectifiers

by   Guodong Zhang, et al.

Training very deep neural networks is still an extremely challenging task. The common solution is to use shortcut connections and normalization layers, which are both crucial ingredients in the popular ResNet architecture. However, there is strong evidence to suggest that ResNets behave more like ensembles of shallower networks than truly deep ones. Recently, it was shown that deep vanilla networks (i.e. networks without normalization layers or shortcut connections) can be trained as fast as ResNets by applying certain transformations to their activation functions. However, this method (called Deep Kernel Shaping) isn't fully compatible with ReLUs, and produces networks that overfit significantly more than ResNets on ImageNet. In this work, we rectify this situation by developing a new type of transformation that is fully compatible with a variant of ReLUs – Leaky ReLUs. We show in experiments that our method, which introduces negligible extra computational cost, achieves validation accuracies with deep vanilla networks that are competitive with ResNets (of the same width/depth), and significantly higher than those obtained with the Edge of Chaos (EOC) method. And unlike with EOC, the validation accuracies we obtain do not get worse with depth.


page 1

page 2

page 3

page 4


Shifting Mean Activation Towards Zero with Bipolar Activation Functions

We propose a simple extension to the ReLU-family of activation functions...

Highway Networks

There is plenty of theoretical and empirical evidence that depth of neur...

Residual Tangent Kernels

A recent body of work has focused on the theoretical study of neural net...

Characterizing signal propagation to close the performance gap in unnormalized ResNets

Batch Normalization is a key component in almost all state-of-the-art im...

Evolving Normalization-Activation Layers

Normalization layers and activation functions are critical components in...

ResNet strikes back: An improved training procedure in timm

The influential Residual Networks designed by He et al. remain the gold-...

Please sign up or login with your details

Forgot password? Click here to reset