Distance-Based Regularisation of Deep Networks for Fine-Tuning
This addresses the challenge of improving generalization in transfer learning for deep networks, offering a theoretically grounded and effective solution for practitioners.
The paper tackles the problem of overfitting during fine-tuning of deep neural networks by proposing a distance-based regularization method that constrains weight updates to a small sphere around pre-trained weights, leading to provably better generalization and outperforming state-of-the-art fine-tuning competitors.
We investigate approaches to regularisation during fine-tuning of deep neural networks. First we provide a neural network generalisation bound based on Rademacher complexity that uses the distance the weights have moved from their initial values. This bound has no direct dependence on the number of weights and compares favourably to other bounds when applied to convolutional networks. Our bound is highly relevant for fine-tuning, because providing a network with a good initialisation based on transfer learning means that learning can modify the weights less, and hence achieve tighter generalisation. Inspired by this, we develop a simple yet effective fine-tuning algorithm that constrains the hypothesis class to a small sphere centred on the initial pre-trained weights, thus obtaining provably better generalisation performance than conventional transfer learning. Empirical evaluation shows that our algorithm works well, corroborating our theoretical results. It outperforms both state of the art fine-tuning competitors, and penalty-based alternatives that we show do not directly constrain the radius of the search space.