Explicit Inductive Bias for Transfer Learning with Convolutional Networks

TL;DR

This paper explores regularization techniques that explicitly encourage similarity to pre-trained models during transfer learning with convolutional networks, improving fine-tuning effectiveness.

Contribution

It introduces and evaluates regularization schemes, especially an $L^2$ penalty, to explicitly incorporate inductive bias towards the pre-trained model in transfer learning.

Findings

Explicit bias improves transfer learning performance

L2 penalty with pre-trained model as reference is effective

Regularization schemes outperform standard fine-tuning methods

Abstract

In inductive transfer learning, fine-tuning pre-trained convolutional networks substantially outperforms training from scratch. When using fine-tuning, the underlying assumption is that the pre-trained model extracts generic features, which are at least partially relevant for solving the target task, but would be difficult to extract from the limited amount of data available on the target task. However, besides the initialization with the pre-trained model and the early stopping, there is no mechanism in fine-tuning for retaining the features learned on the source task. In this paper, we investigate several regularization schemes that explicitly promote the similarity of the final solution with the initial model. We show the benefit of having an explicit inductive bias towards the initial model, and we eventually recommend a simple $L^2$ penalty with the pre-trained model being a reference as the baseline of penalty for transfer learning tasks.