Python metric learning code issue
2 comments
Why MATLAB? If you're looking to optimize, that's the opposite direction I'd be going. Numba, Cython or writing C or Fortran if you really need the speed is where you should be going imo.
You're already using numpy, so I don't think you'll get much of an increase (if any) by porting to Matlab.
Well, based in the content of this link http://wiki.scipy.org/PerformancePython, now I believe there is not that much difference in performance, but I still would like to try the Matlab code
https://gist.github.com/mikebotazzo/2c7931586d1d8a32dd7b