[mlpack-git] [mlpack] Mean shift clustering (#388)

[Ryan Curtin]

> +      arma::Col<double> mhVector = newCentroid - allCentroids.col(i);
> +      
> +      // update the centroid.
> +      allCentroids.col(i) = newCentroid;
> +      
> +      if (arma::norm(mhVector, 2) < stopThresh ||
> +          completedIterations > maxIterations) {
> +        break;
> +      }
> +      
> +    }
> +    
> +  }
> +  
> +  // remove duplicate centroids.
> +  for (size_t i = 0; i < allCentroids.n_cols; ++i) {

Why not check if there is a duplicate centroid once you finish finding the centroid for an individual point?  If you refactor the code to only consider one centroid at a time and only add it to the results if it is not duplicated in the existing list of results, I think you will get faster (and more memory-efficient) code.

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/388/files#r23584017
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20150126/8b1368d0/attachment.html>