Deep Learning Character Recognition C + + program (based on RBM)

Links in http://download.csdn.net/detail/lucky_greenegg/5413211

The code is based on the DBN-RBM character recognition of the MATLAB program written in C + + version,

HTTP://PAN.BAIDU.COM/S/1MGZIGPQ (There are many people say code comments too little, in order to facilitate understanding can first look at the MATLAB code, relatively short, the data is also converted from the inside, and the MATLAB code can be seen directly after the image results, Character image for prediction error)
The test data inside has been converted to TXT, where each row represents an image. Complete
Handwriting character recognition, 5,000 samples, 10 classes 1000 tests. The correct rate is 94%.
Implemented by vs2008+opencv2.4.3. Find yourself writing a program has not been out of OPENCV.
Because I am purely rookie, so the written procedures can only be said to be completed
function, what kind of coding style ah, what robustness ah, readability Ah still can't talk about.
After testing, it should be able to run correctly, the results of the same as MATLAB.
In addition, I really do not want to collect points, but I also poor jingle, so everyone is not easy.

Take your own data test, if you want to effect good, you need to modify the parameters ...

Please refer to "A practical Guide to training restricted Boltzmann machines"

Operation Effect: