Growing a Brain: Fine-Tuning by Increasing Model Capacity

TL;DR

This paper demonstrates that increasing a CNN's capacity by adding units and properly normalizing them during fine-tuning significantly improves transfer learning performance, outperforming traditional fixed-size fine-tuning methods.

Contribution

The authors introduce a method of growing CNNs during fine-tuning by adding units and normalizing them, leading to better adaptation and state-of-the-art results.

Findings

Growing CNNs enhances transfer learning performance.

Proper normalization of new units is crucial for effective growth.

State-of-the-art results achieved on benchmark datasets.

Abstract

CNNs have made an undeniable impact on computer vision through the ability to learn high-capacity models with large annotated training sets. One of their remarkable properties is the ability to transfer knowledge from a large source dataset to a (typically smaller) target dataset. This is usually accomplished through fine-tuning a fixed-size network on new target data. Indeed, virtually every contemporary visual recognition system makes use of fine-tuning to transfer knowledge from ImageNet. In this work, we analyze what components and parameters change during fine-tuning, and discover that increasing model capacity allows for more natural model adaptation through fine-tuning. By making an analogy to developmental learning, we demonstrate that "growing" a CNN with additional units, either by widening existing layers or deepening the overall network, significantly outperforms classic fine-tuning approaches. But in order to properly grow a network, we show that newly-added units must be appropriately normalized to allow for a pace of learning that is consistent with existing units. We empirically validate our approach on several benchmark datasets, producing state-of-the-art results.