Exploiting Invariance in Training Deep Neural Networks

TL;DR

This paper introduces a feature transform technique inspired by animal visual systems that enforces invariance properties in deep neural network training, leading to faster convergence, less parameter tuning, and better generalization across tasks.

Contribution

The paper proposes a novel invariance-based feature transform method that improves training efficiency and generalization in deep neural networks, applicable to vision and language tasks.

Findings

Requires fewer training iterations than baselines.

Achieves superior performance across multiple tasks.

Computational overhead is only 5% of convolution layer operations.

Abstract

Inspired by two basic mechanisms in animal visual systems, we introduce a feature transform technique that imposes invariance properties in the training of deep neural networks. The resulting algorithm requires less parameter tuning, trains well with an initial learning rate 1.0, and easily generalizes to different tasks. We enforce scale invariance with local statistics in the data to align similar samples at diverse scales. To accelerate convergence, we enforce a GL(n)-invariance property with global statistics extracted from a batch such that the gradient descent solution should remain invariant under basis change. Profiling analysis shows our proposed modifications takes 5% of the computations of the underlying convolution layer. Tested on convolutional networks and transformer networks, our proposed technique requires fewer iterations to train, surpasses all baselines by a large margin, seamlessly works on both small and large batch size training, and applies to different computer vision and language tasks.