ARIA: On the Interaction Between Architectures, Initialization and Aggregation Methods for Federated Visual Classification
This work addresses the problem of optimizing federated learning systems for medical image classification, providing insights for researchers and practitioners, though it is incremental in nature.
The study investigated the combined effects of architecture, initialization, and aggregation methods in federated learning for medical image classification, finding that these elements must be selected together to optimize performance, with results indicating specific choices based on task characteristics.
Federated Learning (FL) is a collaborative training paradigm that allows for privacy-preserving learning of cross-institutional models by eliminating the exchange of sensitive data and instead relying on the exchange of model parameters between the clients and a server. Despite individual studies on how client models are aggregated, and, more recently, on the benefits of ImageNet pre-training, there is a lack of understanding of the effect the architecture chosen for the federation has, and of how the aforementioned elements interconnect. To this end, we conduct the first joint ARchitecture-Initialization-Aggregation study and benchmark ARIAs across a range of medical image classification tasks. We find that, contrary to current practices, ARIA elements have to be chosen together to achieve the best possible performance. Our results also shed light on good choices for each element depending on the task, the effect of normalisation layers, and the utility of SSL pre-training, pointing to potential directions for designing FL-specific architectures and training pipelines.