Today I finally release Multiple Back-Propagation (MBP) software under open source (GNU General Public License v3).
MBP allows the training neural networks with the Back-Propagation and the Multiple Back-Propagation algorithms. The source code includes a CUDA implementation of the referred algorithms. Depending on the size of the datasets and on the GPU you can get speedups of over 175x (on a GTX 280).
I’m looking for people interested in improving Multiple Back-Propagation as I don’t have the time needed to implement all the features requested. Also I’m sure this great software can benefit from new ideas and opinions. Please contact me if you want to participate in this project. Your contribution is certainly most welcome.
I think you have done a good job here. Well I want to design a Character recognition system. I am giving the detail here;
Character Recognition
It is often useful to have a machine perform pattern recognition. In particular, machines that can read symbols are very cost effective. A machine that reads banking checks can process many more checks than a human being in the same time. This kind of application saves time and money, and eliminates the requirement that a human perform such a repetitive task. The demonstration appcr1 shows how character recognition can be done with a backpropagation network.
Problem Statement
A network is to be designed and trained to recognize the 26 letters of the alphabet. An imaging system that digitizes each letter cantered in the system’s field of vision is available. The result is that each letter is represented as a 5 by 7 grid of Boolean values.
I want to do the training here using backpropagation. Do you think I can solve this problem using your implementation. I have not seen your code as of now, so shall comment on it after looking at it.
Thanks anyway for this work.
Sorry for editing my post. I had a look at the code. By the way I am looking at the code which is present : here
If you are interested only in the CUDA code (and not in the remaining code of the application) you can find a better documented implementation here (I start this project recently):
Hi Noel
I find your gpumlib project very interesting. I am currently trying to parallelize a backpropagation algorithm using CUDA too, though, so far with rather slow results.
It would be very nice if you could provide some basic compilation instructions for your code or whatever makefile you use for it. Currently I’m not able to compile your BP example at all using my own methods.
And thanks for sharing this stuff, that’s really nice! :)
Hi Noel
I find your gpumlib project very interesting. I am currently trying to parallelize a backpropagation algorithm using CUDA too, though, so far with rather slow results.
It would be very nice if you could provide some basic compilation instructions for your code or whatever makefile you use for it. Currently I’m not able to compile your BP example at all using my own methods.
And thanks for sharing this stuff, that’s really nice! :)