Module parameter_learning 

Parameter learning algorithms for probabilistic graphical models.

This module provides algorithms for learning model parameters from data:

• Maximum Likelihood Estimation (MLE): For fully observed data
• Expectation-Maximization (EM): For partially observed data
• Bayesian Estimation: With Dirichlet priors
• Baum-Welch Algorithm: Specialized EM for Hidden Markov Models

Overview

Parameter learning is the process of estimating the parameters (probabilities) of a probabilistic model from observed data. This module supports both:

• Complete data: All variables are observed (use MLE)
• Incomplete data: Some variables are hidden (use EM)

Examples

// Learn HMM parameters from observed sequences
let mut hmm = HiddenMarkovModel::new(2, 2);
let learner = BaumWelchLearner::new(100, 1e-4);
learner.learn(&mut hmm, &observation_sequences)?;

Modules

utils

Utilities for parameter learning.

Structs

BaumWelchLearner

Baum-Welch algorithm for learning HMM parameters.

BayesianEstimator

Bayesian parameter estimator with Dirichlet priors.

MaximumLikelihoodEstimator

Maximum Likelihood Estimator for discrete distributions.

SimpleHMM

Simple HMM representation for parameter learning.