soft_lr: Partially Supervised Logistic Regression with Soft Labels...
In bbuchsbaum/discursive: What the Package Does (One Line, Title Case)

soft_lr

R Documentation

Partially Supervised Logistic Regression with Soft Labels (Robust Version)

Description

Fits a logistic regression model using E²M with soft labels, adding robustness: multiple initial attempts, ridge penalty, stable line search, and optional verbosity.

Usage

soft_lr(
  X,
  PL,
  max_iter = 100,
  tol = 1e-06,
  lambda = 1e-05,
  n_starts = 3,
  verbose = FALSE
)

Arguments

`X`	A numeric matrix n x d.
`PL`	A numeric matrix n x K of plausibilities.
`max_iter`	Integer, max E²M iterations. Default 100.
`tol`	Tolerance for convergence in log-likelihood. Default 1e-6.
`lambda`	Ridge regularization parameter for coefficients. Default 1e-5.
`n_starts`	Number of initializations to try. Default 3.
`verbose`	Logical, if TRUE prints progress. Default FALSE.

Value

A list:

`beta`	(d+1) x (K-1) coefficient matrix
`zeta`	Posterior class probabilities n x K
`loglik`	Final log-likelihood
`iter`	Iterations

References

Quost, B., Denoeux, T., Li, S. (2017). Parametric classification with soft labels using the Evidential EM algorithm. Advances in Data Analysis and Classification, 11(4), 659-690.

bbuchsbaum/discursive documentation built on April 14, 2025, 4:57 p.m.