[mlpack-git] [mlpack/mlpack] GammaDistribution: Adds functionality to solve #749 (#751)

[Ryan Curtin]

> @@ -53,6 +64,13 @@ void GammaDistribution::Train(const arma::mat& rdata, const double tol)
>    Train(logMeanxVec, meanLogxVec, meanxVec, tol);
>  }
>  
> +// Fits an alpha and beta parameter according to observation probabilities.
> +void GammaDistribution::Train(const arma::mat& observations, 
> +                              const arma::vec& probabilities,
> +                              const double tol)
> +{

Hm, maybe that will work; I haven't thought about it too much.  I think the only issue might be that if you take unlikely points to have low probability, then when you train, this biases the training points towards the high-PDF parts of the distribution, possibly giving (I think) a trained distribution with different properties than the original.  If it was a one-dimensional Gaussian we were training, this would result in a lower variance, but I don't have the intuition to say what it will do the Gamma distribution, only that I think you'll end up with a different Gamma distribution than the one you are taking random samples from.  Hopefully what I've written here is at least somewhat coherent. :)

---
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/751/files/9cb117f671f55186baddf38ce71107a2a3ae027f#r73803620
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160807/fb43a239/attachment.html>