Restoring Neural Network Plasticity for Faster Transfer Learning

TL;DR

This paper introduces a targeted weight re-initialization method to restore neural plasticity in pretrained models, significantly improving transfer learning efficiency and accuracy for CNNs and ViTs with minimal extra computation.

Contribution

It proposes a novel re-initialization strategy to address plasticity loss in transfer learning, enhancing adaptation and performance across models and datasets.

Findings

Improved test accuracy on multiple image classification benchmarks.

Faster convergence during fine-tuning.

Negligible additional computational cost.

Abstract

Transfer learning with models pretrained on ImageNet has become a standard practice in computer vision. Transfer learning refers to fine-tuning pretrained weights of a neural network on a downstream task, typically unrelated to ImageNet. However, pretrained weights can become saturated and may yield insignificant gradients, failing to adapt to the downstream task. This hinders the ability of the model to train effectively, and is commonly referred to as loss of neural plasticity. Loss of plasticity may prevent the model from fully adapting to the target domain, especially when the downstream dataset is atypical in nature. While this issue has been widely explored in continual learning, it remains relatively understudied in the context of transfer learning. In this work, we propose the use of a targeted weight re-initialization strategy to restore neural plasticity prior to fine-tuning. Our experiments show that both convolutional neural networks (CNNs) and vision transformers (ViTs) benefit from this approach, yielding higher test accuracy with faster convergence on several image classification benchmarks. Our method introduces negligible computational overhead and is compatible with common transfer learning pipelines.