AI Paper Atlas

Research map / Optimization / Optimizers

Automatic and Harmless Regularization with Constrained and Lexicographic Optimization: A Dynamic Barrier Approach

Chengyue Gong, Xingchao Liu, qiang liu

NeurIPS 2021 · Optimization › Optimizers · approach: Differentiating through embedded optimization · 3 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 tuned-lambda L2 regularization with LexicoEffNet-B0 / AlphaNet-A0 fine-tuning · Oxford Flower (AlphaNet-A0)top-1 accuracy: 97.7±0.2 -> 97.6±0.1No clear effect
Replace tuned-lambda L2 regularization with LexicoEffNet-B0 · Oxford Flowertop-1 accuracy: 97.2±0.1 -> 97.1±0.1No clear effect
Replace tuned-lambda L2 regularization with LexicoAlphaNet-A0 · Stanford Cartop-1 accuracy: 91.0±0.1 -> 91.6±0.1Improved
Replace linearly-combined UDA loss with LexicoUDA · CIFAR-10 100 labelstest accuracy (%): 75.28±2.39 -> 77.61±1.83Improved
Replace linearly-combined UDA loss with LexicoUDA · CIFAR-10 250 labelstest accuracy (%): 91.18±1.08 -> 91.86±0.85Improved
Replace linearly-combined UDA loss with LexicoUDA · CIFAR-10 1000 labelstest accuracy (%): 94.56±0.21 -> 94.53±0.15No clear effect
Replace linearly-combined UDA loss with LexicoUDA · CIFAR-100 1000 labelstest accuracy (%): 54.85±0.83 -> 56.91±0.78Improved
Replace linearly-combined UDA loss with LexicoUDA · CIFAR-100 2500 labelstest accuracy (%): 66.37±0.28 -> 67.06±0.26Improved
Replace linearly-combined UDA loss with LexicoUDA · CIFAR-100 10000 labelstest accuracy (%): 67.42±0.22 -> 67.57±0.14No clear effect
Replace linear combination (best lambda 5) with LexicoSwin-Base UperNet · Cityscapes valmIoU: 79.6±0.2 -> 79.7±0.1Improved
Replace linear combination (best lambda 10) with LexicoSwin-Base FCN auxiliary · Cityscapes valmIoU: 78.3±0.2 -> 78.5±0.1Improved
Dynamic barrier (ours) instead of linear combination for Pareto profilingsparse representation learning · Toy data, BERT-base NG20/IMDbPareto front coverageImproved

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