Speaker: Yanis Bahroun, Flatiron Institute

Title: Exploration & Extension of The Similarity Matching Framework: Feature Learning & Nonlinear Methods & Transformation Learning

Abstract: Similarity matching (SM) is a framework introduced recently for deriving biologically plausible neural networks from objective functions. Three key biological properties associated with these networks are 1) local learning rules, 2) unsupervised learning, and 3) online implementations. In particular, previous work has demonstrated that unconstrained-in-sign SM (USM) and nonnegative SM (NSM) can lead to neural networks (NN) performing linear principal subspace projection (PSP) and clustering. The first objective of this work is to explore the capabilities of existing SM NN for feature learning. The second objective of this work is to SM beyond linear methods and static images. To incorporate nonlinearity, kernel-based versions of SM are proposed and map onto NNs performing nonlinear online clustering and PSP, outperforming traditional methods. Finally, a new SM cost-function is applied to pairs of consecutive images to develop the Transformation NSM algorithm. This is mapped onto a NN that performs motion detection and recapitulates several salient features of the fly visual system. The proposed approach is also applicable to the general problem of transformation learning.