Channel Scaling: A Scale-and-Select Approach for Transfer Learning

TL;DR

This paper introduces a channel-scaling method for transfer learning that efficiently reduces model size by selecting important features, achieving significant parameter reduction without large datasets or fine-tuning.

Contribution

The novel channel-scaling layer approach enables effective feature channel selection in transfer learning, reducing model size while maintaining or improving performance.

Findings

Reduced model parameters by 95%

Achieved superior classification performance

Effective feature channel pruning without large datasets

Abstract

Transfer learning with pre-trained neural networks is a common strategy for training classifiers in medical image analysis. Without proper channel selections, this often results in unnecessarily large models that hinder deployment and explainability. In this paper, we propose a novel approach to efficiently build small and well performing networks by introducing the channel-scaling layers. A channel-scaling layer is attached to each frozen convolutional layer, with the trainable scaling weights inferring the importance of the corresponding feature channels. Unlike the fine-tuning approaches, we maintain the weights of the original channels and large datasets are not required. By imposing L1 regularization and thresholding on the scaling weights, this framework iteratively removes unnecessary feature channels from a pre-trained model. Using an ImageNet pre-trained VGG16 model, we demonstrate the capabilities of the proposed framework on classifying opacity from chest X-ray images. The results show that we can reduce the number of parameters by 95% while delivering a superior performance.