SelfNormalizing Neural Networks
This topic contains 0 replies, has 1 voice, and was last updated by arXiv 1 year ago.

SelfNormalizing Neural Networks
Deep Learning has revolutionized vision via convolutional neural networks (CNNs) and natural language processing via recurrent neural networks (RNNs). However, success stories of Deep Learning with standard feedforward neural networks (FNNs) are rare. FNNs that perform well are typically shallow and, therefore cannot exploit many levels of abstract representations. We introduce selfnormalizing neural networks (SNNs) to enable highlevel abstract representations. While batch normalization requires explicit normalization, neuron activations of SNNs automatically converge towards zero mean and unit variance. The activation function of SNNs are “scaled exponential linear units” (SELUs), which induce selfnormalizing properties. Using the Banach fixedpoint theorem, we prove that activations close to zero mean and unit variance that are propagated through many network layers will converge towards zero mean and unit variance — even under the presence of noise and perturbations. This convergence property of SNNs allows to (1) train deep networks with many layers, (2) employ strong regularization, and (3) to make learning highly robust. Furthermore, for activations not close to unit variance, we prove an upper and lower bound on the variance, thus, vanishing and exploding gradients are impossible. We compared SNNs on (a) 121 tasks from the UCI machine learning repository, on (b) drug discovery benchmarks, and on (c) astronomy tasks with standard FNNs and other machine learning methods such as random forests and support vector machines. SNNs significantly outperformed all competing FNN methods at 121 UCI tasks, outperformed all competing methods at the Tox21 dataset, and set a new record at an astronomy data set. The winning SNN architectures are often very deep. Implementations are available at: github.com/bioinfjku/SNNs.
SelfNormalizing Neural Networks
by Günter Klambauer, Thomas Unterthiner, Andreas Mayr, Sepp Hochreiter
https://arxiv.org/pdf/1706.02515v5.pdf
You must be logged in to reply to this topic.