Deep neural networks with dependent weights: \\Gaussian Process mixture limit, heavy tails, sparsity and compressibility

Hoil Lee, Fadhel Ayed, Paul Jung, Juho Lee, Hongseok Yang, Francois Caron

Keywords: Deep neural network, infinite-width, infinite divisibility, sparsity, compressibility, Gaussian process, triangular arrays, regular variation, pruning

TL;DR: We introduce scalar parameters associated to each node of a feedforward neural net, and discuss the benefits and effects this has on the architecture.

Abstract: This work studies the infinite-width limit of deep feedforward neural networks whose weights are dependent, and modelled via a mixture of Gaussian distributions. Under this model, we show that each layer of the infinite-width neural network can be characterised by two simple quantities: a non-negative scalar parameter and a L\'evy measure on the positive reals. If the scalar parameters are strictly positive and the L\'evy measures are trivial at all hidden layers, then one recovers the classical Gaussian process (GP) limit, obtained with iid Gaussian weights. More interestingly, if the L\'evy measure of at least one layer is non-trivial, we obtain a mixture of Gaussian processes (MoGP) in the large-width limit. The behaviour of the neural network in this regime is very different from the GP regime. One obtains correlated outputs, with non-Gaussian distributions, possibly with heavy tails. We illustrate some of the benefits of the MoGP regime over the GP regime in terms of representation learning and compressibility on simulated, MNIST and Fashion MNIST datasets.

Submission Number: 3