🧠 Deep Learning

Diffusion models, Graph Neural Networks, Multi-task Learning

• AdsorbDiff: Diffusion framework [ICML 2024]
• JMP: Atomic foundation model [ICLR 2024]
• SCN: Novel GNNs [NeurIPS 2022]

⚗️ Catalysis

DFT datasets, analysis, and applications

• OC22 Dataset: Large-scale catalyst datasets [ACS Catalysis 2023]
• Error Analysis: Training metrics [ACS Catalysis 2022]

🌟 AdsorbDiff

ICML 2024 - Diffusion Models for Catalyst-Adsorbate Configurations

Novel diffusion model approach for generating catalyst-adsorbate configurations, enabling efficient exploration of chemical space for catalysis applications.

🔬 JMP

ICLR 2024 - Joint Multi-domain Pretraining

Atomic foundation model trained across molecules, materials, and proteins, demonstrating transfer learning capabilities across diverse 3D atomic systems.

📊 SCN

NeurIPS 2022 - Spherical Channel Networks

Novel graph neural network architecture for large-scale molecular datasets, achieving state-of-the-art performance on Open Catalyst benchmarks.

📈 OC22 Dataset

ACS Catalysis 2023 - Large-scale Catalyst Datasets

Comprehensive dataset for oxide electrocatalysis with over 62M DFT calculations, enabling ML model development for energy applications.

🔄 TAAG

JCP 2022 - Transfer Learning Method

Transfer learning approach for graph neural networks across different molecular domains, improving model generalization and reducing training requirements.

🌀 SpinConv

arXiv 2021 - Novel GNN Architecture

Innovative graph convolution approach using spin-based representations for improved molecular property prediction and catalyst design.