kohonenDTW-package: Supervised and Unsupervised Self-Organising Maps for...

In e-sensing/kohonenDTW: Supervised and Unsupervised Self-Organising Maps for Satellite Image Time Series

Description

Functions to train self-organising maps (SOMs) generated by satellite image time series. Also interrogation of the maps and prediction using trained maps are supported. The name of the package refers to Teuvo Kohonen, the inventor of the SOM.

Details

The kohonen package implements several forms of self-organising maps (SOMs). Online and batch training algorithms are available; batch training can also be done in parallel. Multiple data layers may be presented to the training algorithm, with potentially different distance measures for each layer. The overall distance is a weighted average of the layer distances. Layers may be selected through the whatmap argument, or by providing a weight of zero. The basic function is supersom; som is simply a wrapper for SOMs using just one layer (the classical form).

New data may be mapped to a trained SOM using the map.kohonen function. Function predict.kohonen will map data to the SOM, and will return predictions (i.e., average values for winning units) for those layers that are not in the new data object.

Several visualisation methods are available in function plot.kohonen.

Index: This package was not yet installed at build time.

Author(s)

Lorena Alves, based on package "kohonen" by Ron Wehrens and Johannes Kruisselbrink

Maintainer: Lorena Alves <[e-mail]>

e-sensing/kohonenDTW documentation built on May 27, 2019, 3:29 p.m.