gbphate

Generalized Bayesian Clustering with PHATE Geometry

Cluster analysis is one of the most fundamental problems in data science. We provide a variety of algorithms from clustering to the learning on the space of partitions. See Hennig, Meila, and Rocci (2016, ISBN:9781466551886) for general exposition to cluster analysis.

Kisung You

T4cluster

Tools for Cluster Analysis

gbphate function

<dl><dt>data</dt>
<dd>an \((n\times p)\) matrix of row-stacked observations.</dd>
<dt>k</dt>
<dd>the number of clusters (default: 2).</dd>
<dt>...</dt>
<dd>extra parameters including<dl>
<dt>nnbd</dt>
<dd>size of nearest neighborhood (default: 5).</dd></dl></dd><dt>alpha</dt>
<dd>decay parameter for Gaussian kernel exponent (default: 2).</dd><dt>gamma_a</dt>
<dd>hyperparameter of the Gamma prior (default: 0).</dd><dt>gamma_b</dt>
<dd>hyperparameter of the Gamma prior (default: 0).</dd><dt>burn_in</dt>
<dd>number of MCMC samples to be thrown away for burn-in (default: 50).</dd><dt>nsample</dt>
<dd>number of MCMC samples to be kept after burn-in (default: 50).</dd><dt>random.start</dt>
<dd>a logical to use random initialization or not (default: <code>TRUE</code>).</dd><dt>print.progress</dt>
<dd>a logical to show how much an algorithm is proceeded (default: <code>FALSE</code>.)</dd>
</dl>

Arguments

Generalized Bayesian Clustering with PHATE Geometry — gbphate

<dl>

<dt>data</dt>
<dd>an \((n\times p)\) matrix of row-stacked observations.</dd>


<dt>k</dt>
<dd>the number of clusters (default: 2).</dd>


<dt>...</dt>
<dd>extra parameters including<dl>
<dt>nnbd</dt>
<dd>size of nearest neighborhood (default: 5).</dd>

<dt>alpha</dt>
<dd>decay parameter for Gaussian kernel exponent (default: 2).</dd>

<dt>gamma_a</dt>
<dd>hyperparameter of the Gamma prior (default: 0).</dd>

<dt>gamma_b</dt>
<dd>hyperparameter of the Gamma prior (default: 0).</dd>

<dt>burn_in</dt>
<dd>number of MCMC samples to be thrown away for burn-in (default: 50).</dd>

<dt>nsample</dt>
<dd>number of MCMC samples to be kept after burn-in (default: 50).</dd>

<dt>random.start</dt>
<dd>a logical to use random initialization or not (default: <code>TRUE</code>).</dd>

<dt>print.progress</dt>
<dd>a logical to show how much an algorithm is proceeded (default: <code>FALSE</code>.)</dd>


</dl></dd>

</dl>

gbphate: Generalized Bayesian Clustering with PHATE Geometry

Description

Usage

Arguments