eXplainable AI for Scientific Applications

1. Robust Evaluation [Project] [ICLR24] [TrustLOG@WWW] [Code]

We present a family of robust fidelity metrics for explainable graph neural networks to alleviate the OOD(Out-of-Distribution) problem.

2. Architecture & Algorithm Designs

Graphs: [NeurIPS 20], [SIGKDD 23], [TPAMI 24], [AAAI 24], [ICML 24a]
Time Series: [ICML 24b], [ICLR 24c]

3. Applications

Health Informatics: Protein Network, Patient Network
Security Applications: Provenance Graphs
Environmental Science: [Project] Sea Level Variability, Compound Flooding [ICML 24b], Harmful Algae Bloom

Foundation Models for Scientific Applications

1. Information-Theoretic Foundation

Data Augmentation: [Arxiv 24]

2. Self-Supervised Learning

Graphs : [NeurIPS 21]
Time Series: [AAAI 23] [ICLR 24b]

3. Architecture & Algorithm Designs