English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

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

MixMo: Mixing Multiple Inputs for Multiple Outputs via Deep Subnetworks