Well structured and tested Matlab and C++ code with many features implemented.
|6 Dec 2013||Sergey Demyanov||
In addition to the classic CNN capabilities the library supports max-pooling, rectified linear units, dropout, momentum, adjusted learning rates, SVM residuals on the last level, adjustment for unbalances datasets and more.
The fact that both versions work identically insures that the code is free of bugs. C++ version works about 2 times faster, and Matlab version allows to view intermediate values and quickly modify the code without recompilation. The code in both versions is quite clean and easy for understanding. If necessary, new types of the layers may be easily added without touching the most of the code. C++ version might be also extended for CUDA platform.