create_predict_fun

create_predict_fun.Learner

Create the algorithms prediction function.

We present 'FACT' (Feature Attributions for ClusTering), a
framework for unsupervised interpretation methods that can be used with an
arbitrary clustering algorithm. The package is capable of re-assigning instances to
clusters (algorithm agnostic), preserves the integrity of the data and does
not introduce additional models. 'FACT' is inspired by the principles of
model-agnostic interpretation in supervised learning. Therefore, some of the
methods presented are based on 'iml', a R Package for Interpretable Machine
Learning by Christoph Molnar, Giuseppe Casalicchio, and Bernd Bischl (2018)
<doi:10.21105/joss.00786>.

Henri Funk

FACT

Feature Attributions for ClusTering

Christian Scholbeck

Giuseppe Casalicchio

create_predict_fun function

<dl><dt>model</dt>
<dd>any 
An arbitrary trained clustering algorithm.</dd>
<dt>task</dt>
<dd><code>character(1)</code> 
Should be clustering in this case. To be extended...</dd>
<dt>predict.fun</dt>
<dd><code>function</code> 
The function to assign newdata. Only needed if <code>model</code> is not a model from
<code>mlr3cluster</code>. The first argument of <code>predict.fun</code> has to be the model, the
second the <code>newdata</code>:
<div class="sourceCode"></div><pre><code>function(model, newdata)
</code></pre>
To be extended for more methods.</dd>
<dt>type</dt>
<dd><code>character(1)</code> 
For soft label predictions, <code>type="prob"</code>. For hard label predictions,
<code>type="partition"</code>. Consult the documentation or definition of the
clustering algorithm you use to find which type options you have.</dd></dl>

Arguments

<ul>
<li><code>create_predict_fun(Learner)</code>: Create a predict function for algorithms from
<code>mlr3cluster</code></li>
</ul>

Methods (by class)

Create a generic prediction function — create_predict_fun

<dl>

<dt>model</dt>
<dd>any 
An arbitrary trained clustering algorithm.</dd>


<dt>task</dt>
<dd><code>character(1)</code> 
Should be clustering in this case. To be extended...</dd>


<dt>predict.fun</dt>
<dd><code>function</code> 
The function to assign newdata. Only needed if <code>model</code> is not a model from
<code>mlr3cluster</code>. The first argument of <code>predict.fun</code> has to be the model, the
second the <code>newdata</code>:
<div class="sourceCode"><pre><code>function(model, newdata)
</code></pre></div>
To be extended for more methods.</dd>


<dt>type</dt>
<dd><code>character(1)</code> 
For soft label predictions, <code>type="prob"</code>. For hard label predictions,
<code>type="partition"</code>. Consult the documentation or definition of the
clustering algorithm you use to find which type options you have.</dd>

</dl>

create_predict_fun: Create a generic prediction function

Description

Usage

Value

Arguments

Methods (by class)