Multiple Back-Propagation is an open source software application for training neural networks with the backpropagation and the multiple back propagation algorithms. Currently this project is also hosted at http://code.google.com/p/multiplebackpropagation
- Fast training (supports CUDA)
- Generates C code for the trained networks
- Allows input sensitivity analysis
- Highly configurable
- Easy to use
- Provides RMS graphics during training
- Provides Output versus Desired output graphics during training for the training and test sets
Overall pretty good. It works for the most part (read: is usable). Not everything works with CUDA so for some things you'll still need CPU (like mbp networks - they dont train properly on CUDA). Documentation is non existent and some bugs present.
small and efficient and gets the job done.
Hi Noel ! Maybe you already abandon this project ! I have a Problems with CUDA in 2.2.2" on Windows 7 /64 . According to other software's' bugs, it seems that all Fermi architecture and or libraries/gpu are not supported ( GTX 570 ) experience this bug, the design of the Nvidia support 2.0 . Thanks Ispas Aurel
Excellent integration with CUDA, lightening fast NN implementation. Lacks Kohonen,SVM, etc. support anyway, only MLP variants.