Homework in Probabilistic Models of Visual Cortex

Homework 1

Topics:

Hierarchy of visual processing, Linear Filter and edge detection, Laplacian of Gaussian.

All the information is here.

Homework 2

Topics:

Bayesian Decision theory and Signal Detection theory, Bayesian method on Edge detection, Fourier Transformation, Sparse Coding

Many data files are deleted: IMAGE.mat, IMAGE_RAW.mat, patchsin_highfre.mat,patchsin_lowfre.mat. These files can be obtained from sparenet or generated from the code.

Also, figures are deleted. All the figures can be generated from the code.

Homework 3

Topics:

Gibbs distribution, Gibbs Sampling, Mean Field Theory, and their relationship

All information is here.

Homework 4

Topics:

Perceptron, Multi-layer Perceptron, Deep Network

The Trained model is deleted, but it can be reproduced by the lua code. My code did not do a good job in translational invariance and I don't really know why it is the case. There could be a bug I guess.

For the data, unzip the CNN_final.zip