Deep learning with Elastic Averaging SGD

Abstract

We study the problem of stochastic optimization for deep learning in the parallel computing environment under communication constraints. A new algorithm is proposed in this setting where the communication and coordination of work among concurrent processes (local workers), is based on an elastic force which links the parameters they compute with a center variable stored by the parameter server (master). The algorithm enables the local workers to perform more exploration, i.e. the algorithm allows the local variables to fluctuate further from the center variable by reducing the amount of communication between local workers and the master. We empirically demonstrate that in the deep learning setting, due to the existence of many local optima, allowing more exploration can lead to the improved performance. We propose synchronous and asynchronous variants of the new algorithm. We provide the stability analysis of the asynchronous variant in the round-robin scheme and compare it with the more common parallelized method ADMM. We show that the stability of EASGD is guaranteed when a simple stability condition is satisfied, which is not the case for ADMM. We additionally propose the momentum-based version of our algorithm that can be applied in both synchronous and asynchronous settings. Asynchronous variant of the algorithm is applied to train convolutional neural networks for image classification on the CIFAR and ImageNet datasets. Experiments demonstrate that the new algorithm accelerates the training of deep architectures compared to DOWNPOUR and other common baseline approaches and furthermore is very communication efficient.

NIPS2015 camera-ready version

References in corpus (4)

• OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks
• The Loss Surfaces of Multilayer Networks
• Slow Learners are Fast
• GPU Asynchronous Stochastic Gradient Descent to Speed Up Neural Network Training

Cited by in corpus (16)

• Revisiting Distributed Synchronous SGD
• Asynchronous Stochastic Gradient Descent with Delay Compensation
• Scaling Deep Learning on GPU and Knights Landing clusters
• ImageNet Training in Minutes
• FedAT: A High-Performance and Communication-Efficient Federated Learning System with Asynchronous Tiers
• Communication trade-offs for synchronized distributed SGD with large step size
• An MPI-Based Python Framework for Distributed Training with Keras
• Minimax Estimation for Personalized Federated Learning: An Alternative between FedAvg and Local Training?
• Improved Dropout for Shallow and Deep Learning
• Distributed Deep Learning for Question Answering
• EventGraD: Event-Triggered Communication in Parallel Machine Learning
• Partial local entropy and anisotropy in deep weight spaces
• Accelerating Asynchronous Stochastic Gradient Descent for Neural Machine Translation
• Asynchronous Federated Learning for Sensor Data with Concept Drift
• BPGrad: Towards Global Optimality in Deep Learning via Branch and Pruning
• ${\text{A}}^2{\text{CiD}}^2$: Accelerating Asynchronous Communication in Decentralized Deep Learning