ordinalRidge: Ordinal regression with a ridge regularization penalty

In labsyspharm/ordinalRidge: Fast solver for ordinal logistic regression with a ridge penalty

Description

Ordinal regression with a ridge regularization penalty

Usage

ordinalRidge(
  K,
  y,
  kernel = (nrow(K) == ncol(K)),
  lambda = 0.1,
  eps = 1e-05,
  maxIter = 10,
  verbose = TRUE
)

## S3 method for class 'ordinalRidge'
predict(mdl, newdata)

Arguments

K	an n-by-p matrix of n samples in p dimensions or an n-by-n kernel matrix
y	n-by-1 vector of (ordinal) labels
kernel	set TRUE if K is a kernel matrix (Default: TRUE if K is a square matrix, FALSE otherwise)
lambda	regularization coefficient
eps	convergence tolerance
maxIter	maximum number of iterations
verbose	if TRUE, reports objective function value at each iteration
mdl	model returned by ordinalRidge()
newdata	n1-by-n kernel matrix of n1 new points against n points used for training

Value

OrdinalRidge returns a list with the following elements: #'

v

A n-by-1 vector of kernel weights

b

A nb-by-1 vector of bias terms to be used as decision boundaries

classes

Names of classes from the original vector of labels y

predict() returns a list with the following elements:

score

An n1-by-1 vector of scores

pred

A factor of length n1 containing predictions

prob

An n1-by-nb matrix of probabilities for each of the nb decision boundaries

Methods (by generic)

• predict: Predict method for ordinalRidge models

labsyspharm/ordinalRidge documentation built on Dec. 21, 2021, 8:49 a.m.