Research

AI-GENESIS: Advancing Intelligence for Graph-Based Exploration of Novel Electronic Systems

Navigating vast design spaces, intricate performance tradeoffs, resource-intensive simulations, and rapid technological evolution presents substantial hurdles. Our novel Graph-based Semi-Supervised Optimization Framework offers a comprehensive solution to this complex issue. By breaking down the problem into distinct key stages, we achieve both alignment with the actual process and remarkable computational efficiency, without compromising effectiveness.

Gravimeter data involves measuring changes in the strength of the gravitational field across different locations on Earth. A gravitational anomaly indicates deviations from the expected field strength, uncovering variations in the distribution of mass beneath the surface. By leveraging these anomalies, we provide solutions that utilize graph-based machine learning methods to analyze Earth's subsurface characteristics. This approach offers valuable insights into density anomalies, contributing to a range of fields including geophysics, geology, and environmental monitoring.

Plastic pollution of water bodies, such as oceans and rivers, is a global environmental problem. The SPIHSI project leverages AI to identify and track plastic pollution hotspots in water bodies from available satellite imagery.

Modeling Spatio-Temporal processes on road networks is a task of paramount significance. ScAI researchers designed Frugal Spatio-temporal Forecasting on Road Networks (FRIGATE). FRIGATE is powered by a spatio-temporal graph neural network (GNN) that integrates positional, topological, and temporal information click here: https://arxiv.org/abs/2306.08277