Looking for the latest version? Download nnet.zip (9.0 MB)
Home
Name Modified Size Downloads / Week Status
Totals: 6 Items   10.9 MB 30
README 2011-02-22 1.7 kB 22 weekly downloads
Thesis_.doc 2011-02-22 1.9 MB 22 weekly downloads
flowChart1.dia 2011-02-10 5.5 kB 11 weekly downloads
flowChart.dia 2011-02-10 2.5 kB 11 weekly downloads
index.html 2011-02-09 1.0 kB 11 weekly downloads
nnet.zip 2011-02-09 9.0 MB 2323 weekly downloads
Backpropagation Neural network nnet.zip contains two version of code. Element wise contains my own code that create a backpropagation neural network. (Recommended if you wanted to learn Back propagation Neural network. MatlabNet contains the Neural net (NN) code that uses buildin matlab (newff, ... ) function. (recommended because it is optimized) Minimun Requirment: 4GB memory For effiency use 64bit OS. Note: Affine transformations in used maple toolkit that comes with matlab2007. I had used matlab 7 (2007) on a 32bit platform (mean matlab7 32bit was installed). Maple is discarded in 2009 and onward. so to use transformation use matlab2007. First version: Backpropagation code. We feel that backpropagation neural network uses too much memory so code present in the textbook pseudo code was implemented. Great for learning but not so efficient for practical reasons (it is not optimized, because i used nested loops. Even it did not uses too much memory but on the down side it takes too much processor time. (in the nnet.zip -> elementWise -> v1.2 -> trainBgdE.m). Second, matlab code in the nnet.zip -> MatlabNet -> trBPM.m. Initial GUI could be found (Still incomplete) at nnet.zip -> form -> MainGUI.m Please note that there are lot of backup files (automatically created by matlab), plus the directory structure is also is not in a nomenclature that would more appropriate. Please note that there is a path file that is include at top, which contains the paths to some files that program needs for execution and that path file automatically inserts all those paths alongside matlab paths. This was done so that i did not had to cd to all those directories.
Source: README, updated 2011-02-22