学术报告（刘绍昌 7.6）

Title: Noncommutative geometry in application to machine learning

Abstract: It is an interesting observation that neural network in machine learning shares the same starting point as quiver representation theory.  In this talk, I will build an algebro-geometric formulation of a `computing machine' which is well-defined over the moduli space of representations.  An important algebraic ingredient is extending the associative geometry of Connes, Cuntz-Quillen, Ginzburg to near-rings, in order to accommodate non-linear activation functions in neural network.  Furthermore, I will explain a uniformization between spherical, Euclidean and hyperbolic moduli spaces of framed quiver representations.