Speaker: Prof. Jean-Jacques Slotine (Massachusetts Institute of Technology)

Title: Stable adaptation and learning in dynamical systems

Abstract:

While we may soon have AI-based artists or scientists, we are nowhere near autonomous robot plumbers. The human brain still largely outperforms robotic algorithms in most  tasks, using computational elements 7 orders of magnitude slower than their artificial counterparts. Similarly, current large scale machine learning algorithms require millions of examples and close proximity to power plants, compared to the brain's few examples and 20W consumption. We study how modern nonlinear systems tools, such as contraction analysis, virtual dynamical systems, and adaptive nonlinear control can yield quantifiable insights about collective computation and learning in large physical systems and dynamical networks. For instance, we show how stable implicit sparse regularization can be exploited online in adaptive prediction or control to select relevant dynamic models out of plausible physically-based candidates, and how most elementary results on gradient descent and optimization based on convexity can be replaced by much more general results based on Riemannian contraction.

Time permitting, we will discuss a new approach to dense associative memories and transformers directly inspired by astrocyte biology. This may be the first contribution to AI of neuroscience results from the last 50 years.