Local Critic Training for Model-Parallel Learning of Deep Neural Networks

TL;DR

This paper introduces local critic training, a model-parallel learning approach for deep neural networks that reduces training time and memory usage by decoupling layer updates with local critic networks, applicable to CNNs and RNNs.

Contribution

The paper presents a novel local critic training method that enables efficient, decoupled, and convergent model-parallel training for deep neural networks.

Findings

Achieves significant reduction in training time.

Decreases memory consumption per machine.

Ensures convergence to a critical point.

Abstract

In this paper, we propose a novel model-parallel learning method, called local critic training, which trains neural networks using additional modules called local critic networks. The main network is divided into several layer groups and each layer group is updated through error gradients estimated by the corresponding local critic network. We show that the proposed approach successfully decouples the update process of the layer groups for both convolutional neural networks (CNNs) and recurrent neural networks (RNNs). In addition, we demonstrate that the proposed method is guaranteed to converge to a critical point. We also show that trained networks by the proposed method can be used for structural optimization. Experimental results show that our method achieves satisfactory performance, reduces training time greatly, and decreases memory consumption per machine. Code is available at https://github.com/hjdw2/Local-critic-training.