Sparsenet starter kit!
Author: Jack Culpepper
License: BSD
checkgrad.m was written by Carl Rasmussen
sfigure.m was written by Daniel Eaton
This program is related to the paper by Bruno and David:
B. A. Olshausen and D. J. Field. Emergence of simple-cell receptive field
properties by learning a sparse code for natural images. Nature,
381(6583):607-9, jun 1996.
(Bruno is my PhD advisor.)
Assumes:
IMAGES.mat is in ../data/IMAGES.mat
lbfgsb-stewart is in ../lbfgsb-stewart
To get started:
- Run unittest in matlab.
- Use ~jack/bin/gqview to look in the state dir.
To run positive only coefficients:
- Comment out this line in unittest.m:
nb = zeros(1,M); % bound type (none)
For a new dataset:
- Add a case to the switch statement.
For a new optimization procedure:
- Add a case to the switch statement.
For a new model:
- Write a check and objfun for a new inference objfun.
- Write a check and objfun for a new learning objfun.