AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

PhysMLE: Generalizable and Priors-Inclusive Multi-Task Remote Physiological Measurement

Jiyao Wang, Hao Lü, Ange Wang, Xiao ru Yang, Ying-Cong Chen, Dengbo He, Kaishun Wu

TPAMI 2025 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Augmentation and mixture-of-experts generalization · 23 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
add EFRouter instead of softmax gating (ablation)PhysMLE-R · PURE HR (Protocol I)MAE: 6.95 -> 5.43Improved
add EFRouter instead of softmax gating (ablation)PhysMLE-R · HCW RR (Protocol I)MAE: 7.60 -> 5.90Improved
use LoRA experts instead of full-parameter experts (ablation)PhysMLE-R · PURE HR (Protocol I)MAE: 5.45 -> 5.43No clear effect
use LoRA experts instead of full-parameter experts (ablation)PhysMLE-R · HCW HR (Protocol I)MAE: 7.02 -> 7.13Hurt
add PhysMLE layer to backbone instead of plain multi-head base (ablation)ResNet-18 backbone · PURE HR (Protocol I)MAE: 8.74 -> 5.43Improved
add PhysMLE layer to backbone instead of plain multi-head base (ablation)ViT backbone · PURE SpO2 (Protocol I)MAE: 2.45 -> 1.86Improved
add expert orthogonality loss L_OEM (ablation)PhysMLE-R · PURE, HCW, UBFC (Protocol I)MAEImproved
add L_HB, L_BR and L_ASp physiological prior losses (ablation)PhysMLE-R · PURE, HCW, UBFC (Protocol I)MAEImproved
increase number of experts from 3 to 4 (ablation)PhysMLE-R · HCW and PURE (Protocol I)MAE: 34.57 -> 53.43Mixed
raise LoRA rank r from 8 to 16 at alpha 32 (ablation)PhysMLE-R · PURE HR (Protocol I)MAE: 9.48 -> 5.43Improved
raise LoRA rank r from 16 to 32 at alpha 32 (ablation)PhysMLE-R · PURE HR (Protocol I)MAE: 5.43 -> 6.67Hurt
use ViT backbone instead of ResNet-18 (ablation)PhysMLE · PURE SpO2 (Protocol I)MAE: 2.19 -> 1.86Improved

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