Dimensions overall score 4.0

GitHub Code Pulse

No public code linked for this paper yet.

• BVFLMSP improves discrimination performance over the centralized multimodal baseline MultiSurv by up to 0.02 higher C-index across multimodal settings.result 90%
• BVFLMSP integrates differential privacy mechanisms by perturbing client side representations before transmission to provide formal privacy guarantees.method 90%
• BVFLMSP remains robust under strict privacy constraints while providing uncertainty estimates.result 85%
• Most existing multimodal survival models produce single deterministic predictions without indicating model confidence, limiting their reliability.limitation 80%
• BVFLMSP uses a Split Neural Network architecture where clients independently model specific data modalities using Bayesian neural networks.method 90%
• Centralized model training is impractical for multimodal time-to-event prediction when sensitive data is distributed across multiple parties due to privacy constraints.technical 85%
• BVFLMSP effectively includes multimodal data and improves survival prediction over existing baselines.result 85%
• There is a tradeoff between predictive performance and privacy in federated learning under varying privacy budgets across different modality combinations.result 80%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card