RWNN-object

An RWNN-object is a list containing the following:<dl>
 <dt><code>data</code></dt>
<dd>The original data used to estimate the weights.</dd> <dt><code>n_hidden</code></dt>
<dd>The vector of neurons in each layer.</dd> <dt><code>activation</code></dt>
<dd>The vector of the activation functions used in each layer.</dd> <dt><code>lnorm</code></dt>
<dd>The norm used when estimating the output weights.</dd> <dt><code>lambda</code></dt>
<dd>The penalisation constant used when estimating the output weights.</dd> <dt><code>bias</code></dt>
<dd>The <code>TRUE/FALSE</code> bias vectors set by the control function for both hidden layers, and the output layer.</dd> <dt><code>weights</code></dt>
<dd>The weigths of the neural network, split into random (stored in hidden) and estimated (stored in output) weights.</dd> <dt><code>sigma</code></dt>
<dd>The standard deviation of the corresponding linear model.</dd> <dt><code>type</code></dt>
<dd>A string indicating the type of modelling problem.</dd> <dt><code>combined</code></dt>
<dd>A list of two <code>TRUE/FALSE</code> values stating whether the direct links were made to the input, and whether the output of each hidden layer was combined to make the prediction.</dd>
</dl>

Creation, estimation, and prediction of random weight neural networks (RWNN), Schmidt et al. (1992) <doi:10.1109/ICPR.1992.201708>, including popular variants like extreme learning machines, Huang et al. (2006) <doi:10.1016/j.neucom.2005.12.126>, sparse RWNN, Zhang et al. (2019) <doi:10.1016/j.neunet.2019.01.007>, and deep RWNN, Henríquez et al. (2018) <doi:10.1109/IJCNN.2018.8489703>. It further allows for the creation of ensemble RWNNs like bagging RWNN, Sui et al. (2021) <doi:10.1109/ECCE47101.2021.9595113>, boosting RWNN, stacking RWNN, and ensemble deep RWNN, Shi et al. (2021) <doi:10.1016/j.patcog.2021.107978>.

Sc3<b8>ren Vilsen

RWNN

Random Weight Neural Networks

RWNN-object function

An RWNN-object is a list containing the following:<dl>
 <dt><code>data</code></dt>
<dd>The original data used to estimate the weights.</dd>

 <dt><code>n_hidden</code></dt>
<dd>The vector of neurons in each layer.</dd>

 <dt><code>activation</code></dt>
<dd>The vector of the activation functions used in each layer.</dd>

 <dt><code>lnorm</code></dt>
<dd>The norm used when estimating the output weights.</dd>

 <dt><code>lambda</code></dt>
<dd>The penalisation constant used when estimating the output weights.</dd>

 <dt><code>bias</code></dt>
<dd>The <code>TRUE/FALSE</code> bias vectors set by the control function for both hidden layers, and the output layer.</dd>

 <dt><code>weights</code></dt>
<dd>The weigths of the neural network, split into random (stored in hidden) and estimated (stored in output) weights.</dd>

 <dt><code>sigma</code></dt>
<dd>The standard deviation of the corresponding linear model.</dd>

 <dt><code>type</code></dt>
<dd>A string indicating the type of modelling problem.</dd>

 <dt><code>combined</code></dt>
<dd>A list of two <code>TRUE/FALSE</code> values stating whether the direct links were made to the input, and whether the output of each hidden layer was combined to make the prediction.</dd>


</dl>

RWNN-object: An RWNN-object

Description

Arguments