add_neurons: Add hidden neurons to the ELM.

In mugiro/elm: Extreme Learning Machine for regression and classification

Description

add_neurons adds a specific number of hidden neurons to the ELM being all of them of the same type of activation function.

Usage

add_neurons(object, ...)

Arguments

object	An instance to the ELM class.
nn	The number of hidden neurons to add to the network.
act_fun	The activation function of the added neurons. Several types: "linear" A standard linear function. "sigmoid" A mathematical function having an "S" shape. "tan" "rbf"
w_in	An input weight matrix of dimension [dxL]. List of centroids for rbf activation functions.
b	An input bias vector of dimension [1xL]. Vector of sigmas for rbf activation functions.

Value

An object ELM with new neurons added and w_in and b matrices updated.

It is called by the training wrapper when a new ELM object is created. It is called sequentially based on the different type of activation functions.

When addNeurons is called explicitly, the ELM should be re-trained

For linear activation functions, the number of neurons added cannot be superior to the number of features (L=<d). This case entails a linear projection of data to a higher dimensional, which yields a multicorrelated new space.

mugiro/elm documentation built on May 23, 2019, 8:21 a.m.