Federated Learning Over Images: Vertical Decompositions and Pre-Trained Backbones Are Difficult to Beat

TL;DR

This paper evaluates various federated learning algorithms on image classification tasks, highlighting that vertical network decomposition and pre-trained backbones generally outperform other methods across diverse settings.

Contribution

It provides a comprehensive evaluation of federated learning strategies for images, emphasizing the effectiveness of vertical decompositions and pre-trained backbones.

Findings

Vertical decompositions outperform standard methods.

Pre-trained backbones are difficult to beat.

Evaluation metrics beyond accuracy are necessary.

Abstract

We carefully evaluate a number of algorithms for learning in a federated environment, and test their utility for a variety of image classification tasks. We consider many issues that have not been adequately considered before: whether learning over data sets that do not have diverse sets of images affects the results; whether to use a pre-trained feature extraction "backbone"; how to evaluate learner performance (we argue that classification accuracy is not enough), among others. Overall, across a wide variety of settings, we find that vertically decomposing a neural network seems to give the best results, and outperforms more standard reconciliation-used methods.