DMS Statistics and Data Science Seminar

Title:  baller2vec : A Multi-Entity Transformer For Multi-Agent Spatiotemporal Modeling

Abstract: Multi-agent spatiotemporal modeling, e.g., forecasting the trajectories of basketball players during a game, is a challenging task from both an algorithmic design and computational complexity perspective. Recent work has explored the efficacy of traditional deep sequential models in this domain, but these architectures are slow and cumbersome to train, particularly as model size increases. Further, prior attempts to model interactions between agents across time have limitations, such as imposing an order on the agents, or making assumptions about their relationships. In this presentation, I will introduce baller2vec, a multi-entity generalization of the standard Transformer that, with minimal assumptions, can simultaneously and efficiently integrate information across entities and time. To test the effectiveness of baller2vec for multi-agent spatiotemporal modeling, we trained it to perform two different basketball-related tasks: (1) simultaneously forecasting the trajectories of all players on the court and (2) forecasting the trajectory of the ball. Not only does baller2vec learn to perform these tasks well, it also appears to "understand" the game of basketball, encoding idiosyncratic qualities of players in its embeddings, and performing basketball-relevant functions with its attention heads. In addition to discussing some baller2vec results, I will review two fundamental deep learning concepts behind the Transformer architecture: "embeddings" and "attention."

Location: https://auburn.zoom.us/j/83299681626?pwd=SjlGNi9MWWhEMExGM0c0QzBPK0hMZz09