AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

In Defense of the Unitary Scalarization for Deep Multi-Task Learning

Vitaly Kurin, Alessandro De Palma, Ilya Kostrikov, Shimon Whiteson, M. Pawan Kumar

NeurIPS 2022 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 20 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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
replace specialized multi-task optimizers (MGDA, IMTL, PCGrad, GradDrop, RLW) with unitary scalarization plus regularizationMulti-MNIST encoder-decoder · Multi-MNISTavg task test accuracyNo clear effect
unitary scalarization with l2 and dropout instead of specialized optimizersCelebA 40-task · CelebAavg task test accuracyNo clear effect
unitary scalarization instead of specialized optimizersCityscapes · Cityscapesdepth error, segmentation mIOUNo clear effect
stabilize unitary scalarization with larger replay buffer, actor l2, reward normalization at samplingSAC on Meta-World MT10 · Meta-World MT10best average success rateNo clear effect
use PCGrad instead of unitary scalarizationSAC on Meta-World MT50 · Meta-World MT50best average success rateNo clear effect
add explicit regularization (dropout, l2) to unitary scalarization (ablation)unitary scalarization · CelebApeak avg validation accuracyImproved
use specialized optimizers (without explicit regularization) instead of unregularized unitary scalarization (ablation)CelebA · CelebA validationoverfitting onsetImproved
use PCGrad instead of unitary scalarizationCelebA · CelebAtraining time per epoch: roughly 35 times slowerHurt

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