MixMo: Mixing Multiple Inputs for Multiple Outputs via Deep Subnetworks

TL;DR

MixMo introduces a novel framework for multi-input multi-output deep subnetworks that employs feature mixing techniques like CutMix to improve classification performance without additional inference costs.

Contribution

The paper proposes a new mixing mechanism for subnetworks in deep learning, replacing summing with feature mixing inspired by data augmentation, leading to improved accuracy.

Findings

Achieves state-of-the-art results on CIFAR-100 and Tiny ImageNet.

Outperforms data-augmented deep ensembles in accuracy.

Operates efficiently without extra inference or memory overhead.

Abstract

Recent strategies achieved ensembling "for free" by fitting concurrently diverse subnetworks inside a single base network. The main idea during training is that each subnetwork learns to classify only one of the multiple inputs simultaneously provided. However, the question of how to best mix these multiple inputs has not been studied so far. In this paper, we introduce MixMo, a new generalized framework for learning multi-input multi-output deep subnetworks. Our key motivation is to replace the suboptimal summing operation hidden in previous approaches by a more appropriate mixing mechanism. For that purpose, we draw inspiration from successful mixed sample data augmentations. We show that binary mixing in features - particularly with rectangular patches from CutMix - enhances results by making subnetworks stronger and more diverse. We improve state of the art for image classification on CIFAR-100 and Tiny ImageNet datasets. Our easy to implement models notably outperform data augmented deep ensembles, without the inference and memory overheads. As we operate in features and simply better leverage the expressiveness of large networks, we open a new line of research complementary to previous works.