Name
Abstract | ||
|---|---|---|
This paper studies the performance of a recently proposed preconditioned stochastic gradient descent (PSGD) algorithm on recurrent neural network (RNN) training. PSGD adaptively estimates a preconditioner to accelerate gradient descent, and is designed to be simple, general and easy to use, as stochastic gradient descent (SGD). RNNs, especially the ones requiring extremely long term memories, are difficult to train. We have tested PSGD on a set of synthetic pathological RNN learning problems and the real world MNIST handwritten digit recognition task. Experimental results suggest that PSGD is able to achieve highly competitive performance without using any trick like preprocessing, pretraining or parameter tweaking. |
Year | Venue | Field |
|---|---|---|
2016 | arXiv: Machine Learning | Mathematical optimization,Stochastic gradient descent,Gradient descent,MNIST database,Preconditioner,Computer science,Recurrent neural network,Tweaking,Preprocessor,Artificial intelligence,Backpropagation,Machine learning |
DocType | Volume | Citations |
Journal | abs/1606.04449 | 0 |
PageRank | References | Authors |
0.34 | 4 | 1 |