Dr. Manos Papadakis, University of Houston
Mar-14-2023, 10:30 am
Abstract- In this presentation we start with a premise Initiating a learning process for a neural network from zero may not always the best tactic Human brains which are the best we know tools to analyze images evolved over a billion of years and this evolution perfected the visual processing and analysis of neuronal mechanisms In this talk we want to propose a new, simple very customizable design method for the construction of multi dimensional, wavelet like families of affine frames, commonly referred to as framelets with specific directional characteristics, small and compact support in space, and axial symmetries or anti symmetries The framelets we construct arise from readily available refinable functions The filters defining these framelets have few non zero coefficients, and custom selected orientations The filters can act as finite difference operators We also how we have used a special family of framelet filters to register a top 10 score in a neural network competition sponsored by Google in 2019.