Building a Winning Team: Selecting Source Model Ensembles using a Submodular Transferability Estimation Approach
This addresses the challenge of efficiently choosing multiple pre-trained models for ensemble-based transfer learning in computer vision, offering a novel approach to improve performance.
The paper tackles the problem of selecting ensembles of pre-trained models for transfer learning by proposing OSBORN, a metric that accounts for domain differences, task differences, and model cohesiveness, and it outperforms state-of-the-art methods like MS-LEEP and E-LEEP on image classification and semantic segmentation tasks across 28 source datasets, 11 target datasets, 5 model architectures, and 2 pre-training methods.
Estimating the transferability of publicly available pretrained models to a target task has assumed an important place for transfer learning tasks in recent years. Existing efforts propose metrics that allow a user to choose one model from a pool of pre-trained models without having to fine-tune each model individually and identify one explicitly. With the growth in the number of available pre-trained models and the popularity of model ensembles, it also becomes essential to study the transferability of multiple-source models for a given target task. The few existing efforts study transferability in such multi-source ensemble settings using just the outputs of the classification layer and neglect possible domain or task mismatch. Moreover, they overlook the most important factor while selecting the source models, viz., the cohesiveness factor between them, which can impact the performance and confidence in the prediction of the ensemble. To address these gaps, we propose a novel Optimal tranSport-based suBmOdular tRaNsferability metric (OSBORN) to estimate the transferability of an ensemble of models to a downstream task. OSBORN collectively accounts for image domain difference, task difference, and cohesiveness of models in the ensemble to provide reliable estimates of transferability. We gauge the performance of OSBORN on both image classification and semantic segmentation tasks. Our setup includes 28 source datasets, 11 target datasets, 5 model architectures, and 2 pre-training methods. We benchmark our method against current state-of-the-art metrics MS-LEEP and E-LEEP, and outperform them consistently using the proposed approach.