Author:
Refinetti Maria,d’Ascoli Stéphane,Ohana Ruben,Goldt Sebastian
Abstract
Abstract
Direct feedback alignment (DFA) is emerging as an efficient and biologically plausible alternative to backpropagation for training deep neural networks. Despite relying on random feedback weights for the backward pass, DFA successfully trains state-of-the-art models such as transformers. On the other hand, it notoriously fails to train convolutional networks. An understanding of the inner workings of DFA to explain these diverging results remains elusive. Here, we propose a theory of feedback alignment algorithms. We first show that learning in shallow networks proceeds in two steps: an alignment phase, where the model adapts its weights to align the approximate gradient with the true gradient of the loss function, is followed by a memorisation phase, where the model focuses on fitting the data. This two-step process has a degeneracy breaking effect: out of all the low-loss solutions in the landscape, a network trained with DFA naturally converges to the solution which maximises gradient alignment. We also identify a key quantity underlying alignment in deep linear networks: the conditioning of the alignment matrices. The latter enables a detailed understanding of the impact of data structure on alignment, and suggests a simple explanation for the well-known failure of DFA to train convolutional neural networks. Numerical experiments on MNIST and CIFAR10 clearly demonstrate degeneracy breaking in deep non-linear networks and show that the align-then-memorize process occurs sequentially from the bottom layers of the network to the top.
Subject
Statistics, Probability and Uncertainty,Statistics and Probability,Statistical and Nonlinear Physics
Reference48 articles.
1. High-dimensional dynamics of generalization error in neural networks;Advani;Neural Netw.,2020
2. The committee machine: computational to statistical gaps in learning a two-layers neural network;Aubin,2018
3. Statistical mechanics of deep learning;Bahri;Annu. Rev. Condens. Matter Phys.,2020
4. Comparing dynamics: deep neural networks versus glassy systems;Baity-Jesi,2018
5. Neural networks and principal component analysis: learning from examples without local minima;Baldi;Neural Netw.,1989
Cited by
5 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献