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

[Ryan Curtin]

> +{
> + public:
> +  /**
> +   * Create a Mean Shift object and set the parameters which Mean Shift
> +   * will be run with.
> +   *
> +   * @param duplicateThresh If distance of two centroids is less than it, one will be removed.
> +   * @param maxIterations Maximum number of iterations allowed before giving up
> +   * @param stopThresh If the 2-norm of the mean shift vector is less than stopThresh, 
> +   *        iterations will terminate.
> +   * @param kernel Optional KernelType object.
> +   */
> +  MeanShift(const double duplicateThresh = 1.0,
> +            const size_t maxIterations = 1000,
> +            const double stopThresh = 1e-3,
> +            const KernelType kernel = KernelType());

I don't think that the duplicate threshold should have a default, because the value that should be used is completely dependent on the dataset.  I don't mind putting in some basic logic to estimate a threshold value if the user does not specify one, but a default of 1.0 is a bad idea, in my opinion.  (If you think I'm wrong, please feel free to provide a rebuttal.)

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/388/files#r23583442
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20150126/80537169/attachment-0001.html>