Biologically Plausible Training Mechanisms for Self-Supervised Learning in Deep Networks

TL;DR

This paper introduces biologically plausible training methods for self-supervised learning in deep networks, avoiding complex computations and symmetric connections, and demonstrates their effectiveness with CNNs.

Contribution

It proposes a simple, local, contrastive hinge loss and two alternative learning algorithms, DTP and LL, that align with biological plausibility and match standard backpropagation performance.

Findings

Proposed methods achieve comparable accuracy to backpropagation on CNNs.

Contrastive hinge loss simplifies local computations for SSL.

Biologically plausible algorithms enable effective deep network training.

Abstract

We develop biologically plausible training mechanisms for self-supervised learning (SSL) in deep networks. Specifically, by biological plausible training we mean (i) All updates of weights are based on current activities of pre-synaptic units and current, or activity retrieved from short term memory of post synaptic units, including at the top-most error computing layer, (ii) Complex computations such as normalization, inner products and division are avoided (iii) Asymmetric connections between units, (iv) Most learning is carried out in an unsupervised manner. SSL with a contrastive loss satisfies the third condition as it does not require labelled data and it introduces robustness to observed perturbations of objects, which occur naturally as objects or observer move in 3d and with variable lighting over time. We propose a contrastive hinge based loss whose error involves simple local computations satisfying (ii), as opposed to the standard contrastive losses employed in the literature, which do not lend themselves easily to implementation in a network architecture due to complex computations involving ratios and inner products. Furthermore we show that learning can be performed with one of two more plausible alternatives to backpropagation that satisfy conditions (i) and (ii). The first is difference target propagation (DTP) and the second is layer-wise learning (LL), where each layer is directly connected to a layer computing the loss error. Both methods represent alternatives to the symmetric weight issue of backpropagation. By training convolutional neural networks (CNNs) with SSL and DTP, LL, we find that our proposed framework achieves comparable performance to standard BP learning downstream linear classifier evaluation of the learned embeddings.