Resources

Technical Reports

• High-dimensional spectral sampling [report]
• Uncertainty quantification in scientific machine learning [report]

Software

• Spectral Sampling Library: Designing uniform samples in high-dimensions for surrogate modeling and machine learning (coming soon)
• DDxNet: A Multi-Specialty Diagnostic Model for Clinical Time-Series
• GAN priors for unsupervised source separation
• MACC: Manifold and cyclical surrogates for scientific applications
• Function Preserving Projections: A scalable linear projection technique for HD analysis
• GrAMME: Semi-Supervised Learning using Multi-layered Graph Attention Models
• Deep metric learning pipelines for speaker diarization

Presentations

• Designing calibrated uncertainty estimators for deep models, DLCT (ML Collective) [Slides]
• Improving reliability and interpretability of AI models in clinical diagnosis, LLNL DSI AI in Healthcare Workshop [Slides]
• Building reliable and generalizable deep models using prediction calibration [Slides]
• Thoughts on building machine learning models with scientific data [Slides]
• Explaining real-word shifts and explaining under shifts [Slides]
• Understanding behavior of clinical models under domain shifts [Slides]
• Tutorial on unsupervised learning [Part 1] [Part 2] [Part 3]
• Teaching machines to be intelligent: Pitfalls, Challenges and Opportunities [Slides]

Posters

• Research overview of the CASC Machine Intelligence group [Poster]
• A spectral approach for the design of experiments in high dimensions [Poster]
• Designing Inverses for History Matching in Scientific Problems [Poster]
• Building Calibrated Deep Models Via Uncertainty Matching [Poster]
• Modeling Human Brain Connectomes using Structured Neural Networks [Poster]
• Understanding Behavior of Clinical Models under Domain Shifts [Poster]
• Improved Deep Embeddings for Inferencing with Multi-Layered Graphs [Poster]
• Unsupervised Feature Selection using a Blue Noise Graph Spectrum [Poster]