Date of Award
Master of Science
Electrical and Computer Engr
Aaron L Robinson
Assuming the brain is a prediction machine, we investigate the arrangement of 2D visual features (Gabor filters) such that visual prediction can be carried out efficiently. The contributions of this thesis are threefold: (1) we prove that a 1-mode tensor can be predicted from an n-mode and an (n-1)-mode tensors; (2) using a linear model to predict visual features at varying distances from a given location, we show that prediction error increases smoothly and monotonically with increase in distance and increase in degree of sparsity of the prediction model; (3) we show that the visual features can be arranged in a 2D grid such that the total wiring length can be minimized without compromising prediction accuracy. All experiments are carried out using 110 high-resolution natural images.
dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.
Shah, Monika, "On the Arrangement of Visual Features for Efficient and Accurate Prediction using a Linear Model" (2020). Electronic Theses and Dissertations. 2145.