Use SVM model trained in Matlab for classification in python

Question

I have a SVM model trained in MATLAB (using 6 features) for which I have:

• Support Vectors [337 x 6]
• Alpha [337 x 1]
• Bias
• Kernel Function: @rbf_kernel
• Kernel Function Args = 0.9001
• GroupNames [781 x 1]
• Support Vector Indices [337 x 1]
• Scale Data containing:
  • shift [1 x 6]
  • scale factor [1 x 6]

These above are all data that I am able to load in python.

Now I would like to use this model in python without retraining to perform classification in python. In particular I would like to create a SVM model in python from the support vector generated in MATLAB

Is it possible? How? Any help would be very appreciated! I can't retrain it in python because I don't have the training data (and labels) anymore.

Answer 1

I guess you understand how SVM works, so what I would do is to train the model again in python just on the support vectors you found rather than on all the original training data and the result should remain the same (as if you trained it on the full data), since the support vectors are the "interesting" vectors in the data, that are sitting on the boundaries.