lle: find_weights – R documentation

Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!

find_weights

Calculate weight matrix.

Calculates the weights for every neighbour of x_i.

find_weights(nns, X, m, reg = 2, ss = FALSE, p = 0.5, id = FALSE, v = 0.99)

`nns`	matrix of nearest neighbours using `find_nn_k` or `find_nn_eps`.
`X`	matrix object containing the input data.
`m`	intrinsic dimension of the data. See `lle`.
`reg`	regularisation method. See `lle`.
`ss`	a logical values indicating wheather to perform subset selection. See `lle`.
`p`	amount of data remaining after subset selection. See `lle`.
`id`	a logical values indicating wheather to calculate the intrinsic dimension. See `lle`.
`v`	threshold parameter for intrinsic dimension. See details.

A list containing the following variables:

`X`	input data, can change if subset selection is applied
`weights`	weight matrix. If x_i is neighbour of x_j then -1<weights[i,j]<1, else zero.
`choise`	index vector of kept data while subset selection
`id`	additionally to the (optional) printed intrinsic dimension, the vector of intrinsic dimension for every data point is returned by the function, so that the vector can easily be ploted manually.

data( lle_scurve_data )
	X <- lle_scurve_data
	nns <- find_nn_k( X, 5 )
	weights <- find_weights( nns, X, 2, 2 )

Locally linear embedding

v1.1

GPL-3

Authors

Holger Diedrich, Dr. Markus Abel (Department of Physics, University Potsdam)

Initial release

2012-03-21