AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

Learning Compact Features via In-Training Representation Alignment

Xin Li, Xiangrui Li, Deng Pan, Yao Qiang, Dongxiao Zhu

AAAI 2023 · Optimization › Implicit bias and generalization · approach: Neural tangent kernel analysis · 2 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
add class-conditional MMD alignment between two mini-batchesResnet18 · KMNIST testaccuracy: 95.57 -> 95.79Improved
add class-conditional MMD alignment between two mini-batchesResnet18 · FMNIST testaccuracy: 92.43 -> 92.57Improved
add class-conditional MMD alignment between two mini-batchesResnet18 · CIFAR10 testaccuracy: 92.99 -> 93.70Improved
add class-conditional MMD alignment between two mini-batchesResnet18 · STL10 testaccuracy: 70.88 -> 72.78Improved
add class-conditional MMD alignment between two mini-batchesResnet18 · CIFAR100 testaccuracy: 74.19 -> 74.88Improved
add class-conditional MMD alignment between two mini-batchesResnet-101 · CIFAR100 testaccuracy: 75.85 -> 76.85Improved
add class-conditional MMD alignment between two mini-batchesResnet-18 · ImageNet testaccuracy: 50.01 -> 53.13Improved
add class-conditional MMD alignment between two mini-batchesBert-base · AG's News testaccuracy: 92.10 -> 92.45Improved
add class-conditional MMD alignment between two mini-batchesBert-base · Amazon Full testaccuracy: 58.53 -> 60.08Improved
add class-conditional MMD alignment between two mini-batchesBert-base · Yahoo Answers testaccuracy: 66.60 -> 67.47Improved
add class-conditional MMD alignment between two mini-batchesBert-base · Yelp Full testaccuracy: 61.12 -> 61.68Improved
replace vanilla SGD with center lossBert-base · AG's News testaccuracy: 92.10 -> 91.70Hurt

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