Research map / Federated learning and privacy / Personalized and heterogeneous FL
Cross-Modal Federated Human Activity Recognition via Modality-Agnostic and Modality-Specific Representation Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| use FDARN with disentangled shared and modality-specific representations | cross-modal federated HAR · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy | Helped |
| add a shared activity classifier ψsc (ablation) | FDARN · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy: 33.76(0.7) vs 39.80(0.5); 44.60(2.0) vs 55.60(1.2); 42.86(1.4) vs 52.05(1.4); 35.43(1.2) vs 40.35(0.7) | Helped |
| add a private activity classifier ψpc (ablation) | FDARN · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy: 39.12(0.4) vs 39.80(0.5); 54.00(0.6) vs 55.60(1.2); 50.47(0.4) vs 52.05(1.4); 39.17(0.4) vs 40.35(0.7) | Helped |
| add the altruistic encoder and shared classifier (ablation) | FDARN · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy: 35.96(0.6) vs 39.80(0.5); 49.00(0.6) vs 55.60(1.2); 48.13(0.3) vs 52.05(1.4); 37.27(0.5) vs 40.35(0.7) | Helped |
| add the egocentric encoder and private classifier (ablation) | FDARN · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy: 37.77(0.5) vs 39.80(0.5); 53.20(0.4) vs 55.60(1.2); 49.24(0.5) vs 52.05(1.4); 37.86(0.7) vs 40.35(0.7) | Helped |
| add the modality discriminator ψmd (ablation) | FDARN · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy: 36.10(0.2) vs 39.80(0.5); 51.20(0.4) vs 55.60(1.2); 47.51(0.3) vs 52.05(1.4); 36.12(0.3) vs 40.35(0.7) | Helped |
| add the separation loss Lsep (ablation) | FDARN · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy: 37.24(0.2) vs 39.80(0.5); 50.80(1.1) vs 55.60(1.2); 50.21(0.3) vs 52.05(1.4); 38.38(0.6) vs 40.35(0.7) | Helped |
| add the spreadout regularizer Lsp (ablation) | FDARN · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy: 38.57(0.3) vs 39.80(0.5); 54.80(0.4) vs 55.60(1.2); 51.05(0.5) vs 52.05(1.4); 39.47(0.4) vs 40.35(0.7) | Helped |
| use spherical modality discriminative loss instead of cross-entropy loss (ablation) | FDARN modality discriminator · Epic-Kitchens; Multimodal-EA; Stanford-ECM; Ego-Exo-AR | accuracy: 37.75(0.3) vs 39.80(0.5); 53.60(0.5) vs 55.60(1.2); 49.24(0.3) vs 52.05(1.4); 37.26(0.5) vs 40.35(0.7) | Helped |
Closest papers by meaning
- Limitations in Employing Natural Language Supervision for Sensor-Based Human Activity Recognition - And Ways to Overcome Them (AAAI 2025 · similarity 0.84)
- Generalizable Sensor-Based Activity Recognition via Categorical Concept Invariant Learning (AAAI 2025 · similarity 0.83)
- MoPFormer: Motion-Primitive Transformer for Wearable-Sensor Activity Recognition (NeurIPS 2025 · similarity 0.83)
- HARDVS: Revisiting Human Activity Recognition with Dynamic Vision Sensors (AAAI 2024 · similarity 0.83)
- Decomposing and Fusing Intra- and Inter-Sensor Spatio-Temporal Signal for Multi-Sensor Wearable Human Activity Recognition (AAAI 2025 · similarity 0.83)
- Domain-Informed Label Fusion Surpasses LLMs in Free-Living Activity Classification (Student Abstract) (AAAI 2025 · similarity 0.83)
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.