In the month of June, I have tried out various approaches to make python libraries available in Scilab including PIMS (python integration mechanism in scilab) and the newly started jupyter client-server approach.
While using PIMS for scikit-learn I faced various issues,which led us to follow another approach (Jupyter). Any approach we decided to follow for this project would have required me to convert and transfer python objects to SCILAB context which could be used at a later stage.
To begin with it, Philippe and me started reading through source of various ml models in the scikit-learn repository. It was found that major returned objects are numpy arrays, which can easily be converted to Scilab matrices.
For every linear regression model, there is a coefficient array and an intercept; which I have successfully transferred to scilab.Upon transfering it to scilab, I have used a user defined predict method in scilab, to do test set prediction like python from within scilab. The results are exactly same for both python and scilab predictions. Also the plots compared are the same.