Name
Abstract | ||
|---|---|---|
With the fast development of deep learning, it has become common to learn big neural networks using massive training data. Asynchronous Stochastic Gradient Descent (ASGD) is widely adopted to fulfill this task for its efficiency, which is, however, known to suffer from the problem of delayed gradients. That is, when a local worker adds its gradient to the global model, the global model may have been updated by other workers and this gradient becomes delayed. We propose a novel technology to compensate this delay, so as to make the optimization behavior of ASGD closer to that of sequential SGD. This is achieved by leveraging Taylor expansion of the gradient function and efficient approximation to the Hessian matrix of the loss function. We call the new algorithm Delay Compensated ASGD (DC-ASGD). We evaluated the proposed algorithm on CIFAR-10 and ImageNet datasets, and the experimental results demonstrate that DC-ASGD outperforms both synchronous SGD and asynchronous SGD, and nearly approaches the performance of sequential SGD. |
Year | Venue | Field |
|---|---|---|
2017 | ICML | Training set,Asynchronous communication,Stochastic gradient descent,Computer science,Hessian matrix,Algorithm,Artificial intelligence,Deep learning,Artificial neural network,Machine learning,Global model,Taylor series |
DocType | ISSN | Citations |
Conference | International Conference on Machine Learning. 2017: 4120-4129 | 14 |
PageRank | References | Authors |
0.66 | 21 | 7 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Shuxin Zheng | 1 | 15 | 2.02 |
| Qi Meng | 2 | 18 | 3.44 |
| Taifeng Wang | 3 | 179 | 13.33 |
| Wei Chen | 4 | 23 | 1.84 |
| Nenghai Yu | 5 | 2238 | 183.33 |
| Zhi-Ming Ma | 6 | 227 | 18.26 |
| Tie-yan Liu | 7 | 4662 | 256.32 |