Deep Mean Field Theory: Layerwise Variance and Width Variation as Methods to Control Gradient Explosion

Greg Yang Sam S Schoenholz

Machine Learning mathscidoc:1912.43159

2018.2
A recent line of work has studied the statistical properties of neural networks to great success from a {\it mean field theory} perspective, making and verifying very precise predictions of neural network behavior and test time performance. In this paper, we build upon these works to explore two methods for taming the behaviors of random residual networks (with only fully connected layers and no batchnorm). The first method is {\it width variation (WV)}, ie varying the widths of layers as a function of depth. We show that width decay reduces gradient explosion without affecting the mean forward dynamics of the random network. The second method is {\it variance variation (VV)}, ie changing the initialization variances of weights and biases over depth. We show VV, used appropriately, can reduce gradient explosion of tanh and ReLU resnets from \exp (\Theta (\sqrt L)) and \exp (\Theta (\sqrt L)) respectively to constant \exp (\Theta (\sqrt L)) . A complete phase-diagram is derived for how variance decay affects different dynamics, such as those of gradient and activation norms. In particular, we show the existence of many phase transitions where these dynamics switch between exponential, polynomial, logarithmic, and even constant behaviors. Using the obtained mean field theory, we are able to track surprisingly well how VV at initialization time affects training and test time performance on MNIST after a set number of epochs: the level sets of test/train set accuracies coincide with the level sets of the expectations of certain gradient norms or of metric expressivity (as defined in\cite {yang_meanfield_2017}), a measure of expansion in a random neural network. Based on insights from past works in
No keywords uploaded!
[ Download ] [ 2019-12-21 11:26:38 uploaded by Greg_Yang ] [ 710 downloads ] [ 0 comments ]
@inproceedings{greg2018deep,
  title={Deep Mean Field Theory: Layerwise Variance and Width Variation as Methods to Control Gradient Explosion},
  author={Greg Yang, and Sam S Schoenholz},
  url={http://archive.ymsc.tsinghua.edu.cn/pacm_paperurl/20191221112638710961719},
  year={2018},
}
Greg Yang, and Sam S Schoenholz. Deep Mean Field Theory: Layerwise Variance and Width Variation as Methods to Control Gradient Explosion. 2018. http://archive.ymsc.tsinghua.edu.cn/pacm_paperurl/20191221112638710961719.
Please log in for comment!
 
 
Contact us: office-iccm@tsinghua.edu.cn | Copyright Reserved