<p>In <a href="https://github.com/mlpack/mlpack/pull/751#discussion_r73803620">src/mlpack/core/dists/gamma_distribution.cpp</a>:</p>
<pre style='color:#555'>&gt; @@ -53,6 +64,13 @@ void GammaDistribution::Train(const arma::mat&amp; rdata, const double tol)
&gt;    Train(logMeanxVec, meanLogxVec, meanxVec, tol);
&gt;  }
&gt;  
&gt; +// Fits an alpha and beta parameter according to observation probabilities.
&gt; +void GammaDistribution::Train(const arma::mat&amp; observations, 
&gt; +                              const arma::vec&amp; probabilities,
&gt; +                              const double tol)
&gt; +{
</pre>
<p>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. :)</p>

<p style="font-size:small;-webkit-text-size-adjust:none;color:#666;">&mdash;<br />You are receiving this because you are subscribed to this thread.<br />Reply to this email directly, <a href="https://github.com/mlpack/mlpack/pull/751/files/9cb117f671f55186baddf38ce71107a2a3ae027f#r73803620">view it on GitHub</a>, or <a href="https://github.com/notifications/unsubscribe-auth/AJ4bFFW_rvT07fKFJ1cEweESC69fqDRBks5qdfZOgaJpZM4Jd7IK">mute the thread</a>.<img alt="" height="1" src="https://github.com/notifications/beacon/AJ4bFL1Q55dHdwzA2zzn0t3Gx-g62vpiks5qdfZOgaJpZM4Jd7IK.gif" width="1" /></p>
<div itemscope itemtype="http://schema.org/EmailMessage">
<div itemprop="action" itemscope itemtype="http://schema.org/ViewAction">
  <link itemprop="url" href="https://github.com/mlpack/mlpack/pull/751/files/9cb117f671f55186baddf38ce71107a2a3ae027f#r73803620"></link>
  <meta itemprop="name" content="View Pull Request"></meta>
</div>
<meta itemprop="description" content="View this Pull Request on GitHub"></meta>
</div>

<script type="application/json" data-scope="inboxmarkup">{"api_version":"1.0","publisher":{"api_key":"05dde50f1d1a384dd78767c55493e4bb","name":"GitHub"},"entity":{"external_key":"github/mlpack/mlpack","title":"mlpack/mlpack","subtitle":"GitHub repository","main_image_url":"https://assets-cdn.github.com/images/modules/aws/aws-bg.jpg","avatar_image_url":"https://cloud.githubusercontent.com/assets/143418/15842166/7c72db34-2c0b-11e6-9aed-b52498112777.png","action":{"name":"Open in GitHub","url":"https://github.com/mlpack/mlpack"}},"updates":{"snippets":[{"icon":"PERSON","message":"@rcurtin in #751: 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. :)"}],"action":{"name":"View Pull Request","url":"https://github.com/mlpack/mlpack/pull/751/files/9cb117f671f55186baddf38ce71107a2a3ae027f#r73803620"}}}</script>