Utilizing Excess Resources in Training Neural Networks

TL;DR

This paper introduces KFLO, a kernel filtering method that enhances neural network training by overparameterization without increasing inference complexity, applicable to various architectures and datasets.

Contribution

The paper proposes a novel kernel filtering linear overparameterization technique that improves training performance while maintaining low inference complexity.

Findings

KFLO improves test performance across multiple models.

The approach is compatible with various network architectures.

No additional inference cost is introduced.

Abstract

In this work, we suggest Kernel Filtering Linear Overparameterization (KFLO), where a linear cascade of filtering layers is used during training to improve network performance in test time. We implement this cascade in a kernel filtering fashion, which prevents the trained architecture from becoming unnecessarily deeper. This also allows using our approach with almost any network architecture and let combining the filtering layers into a single layer in test time. Thus, our approach does not add computational complexity during inference. We demonstrate the advantage of KFLO on various network models and datasets in supervised learning.