Upcoming Event: Babuška Forum

The Physics Information Computation of Perennial Machine Learning and Discovery

Chandrajit Bajaj, University of Texas at Austin

Abstract

I shall present a principled framework for perennial machine learning systems capable of continuous, adaptive discovery without expensive retraining, by bridging the thermodynamics of computation with differential variational inference. Grounded in the principle that scientific discovery is a process of data compression, the learning agent is reformulated as a computational Maxwell’s demon that actively reduces the algorithmic entropy of the system through measurement. Crucially, the sustainability of this process is governed by Landauer’s principle: the continuous adaptation to new data necessitates the erasure of outdated information, incurring an irreducible thermodynamic cost.

In practical settings, the perennial learning dynamics is optimized as a continuous-time optimal control problem governed by Pontryagin’s Minimum Principle, where the "work" of inference is minimized against the dissipative cost of forgetting. Stability and convergence are ensured through Casimir energy control within a stochastic Port-Hamiltonian framework, allowing the system to shape the latent energy landscape and conserve structural invariants while actively steering the discovery process. This approach unifies ML algorithms for forward and inverse problems into a reversible, energy-efficient, dissipative Hamiltonian flow, realizing a stable inference engine capable of long-horizon discovery under physical constraints.

Biography

Chandrajit Bajaj  is the director of the Center for Computational Visualization, at the Oden Institute for Computational and Engineering Sciences and a Professor of Computer Science at the University of Texas at Austin.  Bajaj holds the Computational Applied Mathematics Chair in Visualization. He was awarded a distinguished alumnus award from the Indian Institute of Technology, Delhi, (IIT, Delhi). He is also a Fellow of The American Association for the Advancement of Science (AAAS), Fellow of the Association for Computing Machinery (ACM), Fellow of the Institute of Electrical and Electronic Engineers (IEEE), and Fellow of the Society of Industrial and Applied Mathematics (SIAM).