you can find a SVM implementation here in the Active Learning toolbox (multisvm): https://github.com/IPL-UV/altoolbox. It uses the Torch3 library but there are some precompiled exectuables too. Hope it helps!

Yes you can implement machine learning techniques in C/C++ or Java. In fact if you want maximum performance you should implement them in C/C++. Java is more robust but the performances could be lower than C/C++. There are a lot of libraries implementing SVM or NN:

- http://www.csie.ntu.edu.tw/~cjlin/libsvm/

-http://dlib.net/svm_ex.cpp.html

-http://svmlight.joachims.org/

-http://opennn.cimne.com/download.asp

-https://github.com/Artelnics/OpenNN

regards,

Bogdan

Thank you @Devis Tuia @Bogdan Oancea for sharing awesome ideas.I am grateful.

Actually,i have implemented my Machine learning research in MATLAB,also generated the source code and Models in same platform.Is it possible now to tranform that code into  C code or C++ or even java?

Would you like to implement a new machine learning algorithms? I have two of them with many advantages over Backpropagation.

Article Border Pairs Method—constructive MLP learning classification algorithm

Chapter Optimization for multi layer perceptron: Without the gradient

https://stackoverflow.com/questions/64582/c-java-performance-for-neural-networks