Sketchy Empirical Natural Gradient Methods for Deep Learning

Abstract

In this paper, we develop an efficient sketchy empirical natural gradient method (SENG) for large-scale deep learning problems. The empirical Fisher information matrix is usually low-rank since the sampling is only practical on a small amount of data at each iteration. Although the corresponding natural gradient direction lies in a small subspace, both the computational cost and memory requirement are still not tractable due to the high dimensionality. We design randomized techniques for different neural network structures to resolve these challenges. For layers with a reasonable dimension, sketching can be performed on a regularized least squares subproblem. Otherwise, since the gradient is a vectorization of the product between two matrices, we apply sketching on the low-rank approximations of these matrices to compute the most expensive parts. A distributed version of SENG is also developed for extremely large-scale applications. Global convergence to stationary points is established under some mild assumptions and a fast linear convergence is analyzed under the neural tangent kernel (NTK) case. Extensive experiments on convolutional neural networks show the competitiveness of SENG compared with the state-of-the-art methods. On the task ResNet50 with ImageNet-1k, SENG achieves 75.9\% Top-1 testing accuracy within 41 epochs. Experiments on the distributed large-batch training show that the scaling efficiency is quite reasonable.

References in corpus (13)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Deep Learning in Neural Networks: An Overview
• Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
• Optimizing Neural Networks with Kronecker-factored Approximate Curvature
• New insights and perspectives on the natural gradient method
• MLPerf Training Benchmark
• Optimization for deep learning: theory and algorithms
• Fast Convergence of Natural Gradient Descent for Overparameterized Neural Networks
• Shampoo: Preconditioned Stochastic Tensor Optimization
• Sketched Ridge Regression: Optimization Perspective, Statistical Perspective, and Model Averaging
• Practical Gauss-Newton Optimisation for Deep Learning
• SPSD Matrix Approximation vis Column Selection: Theories, Algorithms, and Extensions
• Efficient Subsampled Gauss-Newton and Natural Gradient Methods for Training Neural Networks

Cited by in corpus (2)

• NG+ : A Multi-Step Matrix-Product Natural Gradient Method for Deep Learning
• Enhance Curvature Information by Structured Stochastic Quasi-Newton Methods