Geometric Algebra Transformers: Revolutionizing Geometric Data

Explore the groundbreaking Geometric Algebra Transformer (GATr) in this 30-minute video presentation by Taco Cohen, a machine learning researcher at Qualcomm AI Research. Dive into the world of geometric data and discover how GATr, a versatile neural network architecture, is revolutionizing fields like computer vision, robotics, chemistry, and physics. Learn about GATr's ability to handle diverse geometric data types, its use of projective geometric algebra for efficient representations, and its equivariance with respect to E(3). Witness the performance improvements GATr brings to various geometric problems compared to non-geometric baselines. Gain insights into equivariant networks and geometric deep learning from Taco Cohen's expertise, including his journey from co-founding Scyfer to his achievements in machine learning. This presentation, delivered at the Intelligent Systems Conference 2023 in Amsterdam, offers a glimpse into the future of geometric data processing.