SOM_preprocessing: A MVDA Function
In angy89/MVDA_package: Multi-View Biological Data Analysis

Description Usage Arguments Details Value

Given the matrix of prototype we execute clustering with SOM in bootstrap mode. In each bootstrap iterarion we permuted the dataset and trained a rectangular SOM. We then count how many time each couple of object have been clustered together normalized for the number of time they have been selected together in the dataset. We construct a matrix with this information for each couple of object that is used as distance for a hierarchical clustering with ward method.

1 2	SOM_preprocessing(DB = NULL, xdim = NULL, ydim = NULL, nboot = 100, t = 0.95)

`DB`	is your matrix dataset
`xdim`	is the number of unit on the x-axis
`ydim`	is the number of unit on the y-axis
`nboot`	is the number of iteration for the bootstrap method
`t`	is a threshold for the consensus matrix. value below t are set to 0, and value above t are set to 1.
`nCenters`	is the desidered number of cluster

This function calculate clustering by using rectangular som in bootrapping mode. We construct a similarity matrix of the sample by using SOM clustering results that is the imput for hierarchical clustering.

a list containing three field: som.res is the hierarchical clustering results. clustering is the vector with clustering assignment. center is the matrix with center prototypes.

angy89/MVDA_package documentation built on May 7, 2019, 8:58 p.m.