VCR learns valid contextual representations for incomplete wearable signals via orthogonal disentanglement and missing-aware mixture-of-experts, improving robustness across full and missing-modality settings.
Focal: Contrastive learning for multimodal time-series sensing signals in factorized orthogonal latent space
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
VCR: Learning Valid Contextual Representation for Incomplete Wearable Signals
VCR learns valid contextual representations for incomplete wearable signals via orthogonal disentanglement and missing-aware mixture-of-experts, improving robustness across full and missing-modality settings.