AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Knowledge Cascade: Reverse Knowledge Distillation on Nonparametric Multivariate Functional Estimation

Luyang Fang, Haoran Lu, Yongkai Chen, Wenxuan Zhong, Ping Ma

JMLR 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Rethinking distillation objectives · 0 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
tune smoothing parameters on a subsample and transfer (KCas) instead of full-sample GCVsmoothing spline ANOVA · Simulation 2 scenario 2 d=20, n=20000median time (min): 102.5 -> 8.1Improved
KCas instead of GCV on a random subsample (SUB)smoothing spline density estimation · ESC density estimationrelative log-likelihood: 1.0369 -> 1.0475Improved
KCas instead of full-sample GCVdensity estimation · ESC density estimationrelative log-likelihoodImproved
KCas instead of GAM-based GCVdensity estimation · ESC density estimationrelative log-likelihood: 0.5514 -> 1.0475Improved
KCas instead of SKIPnonparametric regression · SUSY regressionrelative MSE: 0.8252 -> 0.7963Improved
KCas instead of GAMnonparametric regression · OCUP regressionrelative MSE: 4.1504 -> 0.9999Improved
KCas instead of GCV on a subsample (SUB)nonparametric regression · WFRN regressionrelative MSE: 1.3604 -> 1.0434Improved
KCas instead of SKIP when covariates are complex (d=6)nonparametric regression · Simulation 2 scenario 1 d=6log RelMSE: SKIP median log-RelMSE greater than 1; KCas close to GCVImproved
use KCas retuning-free transfer instead of full teacher grid search (Retune)ResNet50 teacher on CIFAR-10 · CIFAR-10 (MobileNetV2 student)time (hours): 121.1 -> 6.9Improved
use ResNet18 as student instead of MobileNetV2KCas hyperparameter transfer · CIFAR-10 teacher accuracyaccuracyImproved
use square-root instead of linear learning-rate scalingKCas deep transfer · CIFAR-10accuracyNo clear effect
KCas instead of cookbook hyperparametersResNet50 teacher on CIFAR-10 · CIFAR-10accuracyImproved

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