calc_struct_weights: PLS calculation of weights

In mortenarendt/StructuralKnowledgeModl: Incorporation of Structural Knowledge in Multivariate Models

Description

Inner engine for PLS calculating weights based on a combined objective, maximising 1) covariance between X and Y, 2) description of feature to feature relationship, 3) Alternative kernel representation of sample-data and 4) with possible sparseness.

Usage

calc_struct_weights(X, Ys, QQ, HH, aH, sumabsw = sqrt(dim(X)[2]),
  niter = 20)

Arguments

X	a (preprocessed) n by px dataset
Ys	a (preprocessed) n by py dataset of responses
QQ	a py by py symmetric feature kernel representing similarities between features.
HH	a px by px symmetric sample kernel representing similarities between samples.
sumabsw	a scalar contraint on the L1 norm of the weights.
niter	= 20 (default) number of iterations

mortenarendt/StructuralKnowledgeModl documentation built on May 21, 2019, 11:42 a.m.