Benchmarking Decoupled Neural Interfaces with Synthetic Gradients

TL;DR

This paper benchmarks decoupled neural interfaces with synthetic gradients, demonstrating that SG-DNI can significantly accelerate learning processes in neural networks while maintaining accuracy.

Contribution

It introduces a speed benchmark comparing SG-DNI with standard neural interfaces, highlighting the efficiency and potential of decoupled neural learning methods.

Findings

SG-DNI is over 3 times faster than traditional methods.

SG-DNI maintains comparable accuracy to backpropagation.

Decoupled neural interfaces enable asynchronous learning in neural networks.

Abstract

Artifical Neural Networks are a particular class of learning systems modeled after biological neural functions with an interesting penchant for Hebbian learning, that is "neurons that wire together, fire together". However, unlike their natural counterparts, artificial neural networks have a close and stringent coupling between the modules of neurons in the network. This coupling or locking imposes upon the network a strict and inflexible structure that prevent layers in the network from updating their weights until a full feed-forward and backward pass has occurred. Such a constraint though may have sufficed for a while, is now no longer feasible in the era of very-large-scale machine learning, coupled with the increased desire for parallelization of the learning process across multiple computing infrastructures. To solve this problem, synthetic gradients (SG) with decoupled neural interfaces (DNI) are introduced as a viable alternative to the backpropagation algorithm. This paper performs a speed benchmark to compare the speed and accuracy capabilities of SG-DNI as opposed to a standard neural interface using multilayer perceptron MLP. SG-DNI shows good promise, in that it not only captures the learning problem, it is also over 3-fold faster due to it asynchronous learning capabilities.