AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

RotoGrad: Gradient Homogenization in Multitask Learning

Adrián Javaloy, Isabel Valera

ICLR 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 3 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
combine gradient-magnitude scaling with task-specific rotationsRotoGrad on NYUv2 · NYUv2mIoU: 38.05 -> 39.32Improved
remove gradient-magnitude scaling (Rotate Only) (ablation)RotoGrad on NYUv2 · NYUv2mIoU: 38.05 -> 39.63Improved
remove task-specific rotations (Scale Only) (ablation)RotoGrad on NYUv2 · NYUv2mIoU: 38.05 -> 38.89Improved
use gradient-magnitude scaling without rotations (ablation)RotoGrad on NYUv2 · NYUv2Abs. depth: 0.53 -> 0.54Hurt
use task-specific rotations without gradient-magnitude scaling (ablation)RotoGrad on NYUv2 · NYUv2Mean relative normal-task improvement: -25.70 -> -6.56Mixed
rotate a smaller feature subspace (ablation)RotoGrad on CIFAR10 · CIFAR10F1-scoreHurt
make the rotation optimizer learn faster relative to the network optimizer (ablation)RotoGrad on multi-MNIST · multi-MNIST sum tasktest errorHurt
reduce the rotation learning rate below the reported sweet spot (ablation)RotoGrad on multi-MNIST · multi-MNIST sum tasktest error: 5e-4No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy