Exploring the parameter reusability of CNN

TL;DR

This paper investigates the conditions under which CNN parameters can be effectively reused across tasks, proposing a method to evaluate reusability based on RMSE and layer performance, thereby improving transfer learning efficiency.

Contribution

It introduces a novel approach to assess CNN parameter reusability at the layer level using RMSE and performance metrics, clarifying the underlying reasons for successful transfer learning.

Findings

Reusing CNN parameters improves target task performance when conditions are met.

RMSE between source and target convolution kernels is a key indicator of reusability.

Layer-wise evaluation enhances transfer learning effectiveness.

Abstract

In recent times, using small data to train networks has become a hot topic in the field of deep learning. Reusing pre-trained parameters is one of the most important strategies to address the issue of semi-supervised and transfer learning. However, the fundamental reason for the success of these methods is still unclear. In this paper, we propose a solution that can not only judge whether a given network is reusable or not based on the performance of reusing convolution kernels but also judge which layers' parameters of the given network can be reused, based on the performance of reusing corresponding parameters and, ultimately, judge whether those parameters are reusable or not in a target task based on the root mean square error (RMSE) of the corresponding convolution kernels. Specifically, we define that the success of a CNN's parameter reuse depends upon two conditions: first, the network is a reusable network; and second, the RMSE between the convolution kernels from the source domain and target domain is small enough. The experimental results demonstrate that the performance of reused parameters applied to target tasks, when these conditions are met, is significantly improved.