Linh Huynh
Dartmouth University
Title
Inference and Adaptive Dynamics on High-Dimensional Optimization Random Landscapes
Abstract
What do spin glasses (a subfield of high-dimensional probability and statistics), evolutionary biology, and artificial intelligence have in common? All involve optimization on rugged landscapes where metastable states pose significant challenges to the search for optima. In this talk, I will discuss inference techniques for high-dimensional data and the analysis of adaptive dynamics in this setting for a series of stochastic models that increase in complexity: from density-dependent homogeneous populations to Lotka–Volterra-type heterogeneous populations to interacting particle systems. These projects collectively contribute to the interplay between microscopic interactions and macroscopic collective behaviors, while also providing insights into the question: can a machine exhibit intelligence equivalent to that of a human?
