Reimagining service delivery in the agentic era with Google Public Sector
cloud.google.com·20h ago
AI Research
arXiv CS.AI·Jun 10·5 min read
LongMoE: Longitudinal Multimodal Learning via Trajectory-Aware Mixture-of-Experts
TL;DR
Multimodal clinical learning struggles with missing data and the evolving nature of patient conditions over time. LongMoE, or Longitudinal Mixture-of-Experts, is a framework that integrates context-aware imputation and trajectory modeling to address these issues.
✦ Why It Matters
Engineers can leverage LongMoE to build more resilient models for clinical data analysis that handle missing information effectively.
Key Takeaways
- Utilize LongMoE for robust clinical data analysis.
- Incorporate context-aware imputation to handle missing data.
- Leverage trajectory-aware modeling for disease progression.
- Apply Sparse MoE routing for tailored patient expert selection.
How It Works
LongMoE combines several advanced techniques: a context-aware imputation module fills in missing data based on available modalities, while an attentional tokenization module captures temporal patterns in patient visits. The trajectory-aware encoder models how a patient's condition evolves over time, and the context-conditioned Sparse MoE routing selects the most relevant experts for each patient based on their specific data context.