The Utility of Feature Reuse: Transfer Learning in Data-Starved Regimes
This work addresses the challenge of deploying computer vision systems in domains with limited data, but it is incremental as it builds on existing transfer learning methods.
The paper tackled the problem of applying transfer learning with deep neural networks in data-starved regimes, such as domains with fewer than 100 labeled samples, and found that overparameterization and feature reuse improve performance for image classifiers, including on out-of-distribution data.
The use of transfer learning with deep neural networks has increasingly become widespread for deploying well-tested computer vision systems to newer domains, especially those with limited datasets. We describe a transfer learning use case for a domain with a data-starved regime, having fewer than 100 labeled target samples. We evaluate the effectiveness of convolutional feature extraction and fine-tuning of overparameterized models with respect to the size of target training data, as well as their generalization performance on data with covariate shift, or out-of-distribution (OOD) data. Our experiments demonstrate that both overparameterization and feature reuse contribute to the successful application of transfer learning in training image classifiers in data-starved regimes. We provide visual explanations to support our findings and conclude that transfer learning enhances the performance of CNN architectures in data-starved regimes.