launcher.cpp File Reference

Main launcher file for MCMC simulation with various Bayesian non-parametric processes. More...

#include "samplers/neal.hpp"
#include "samplers/splitmerge.hpp"
#include "samplers/splitmerge_SAMS.hpp"
#include "processes/DP.hpp"
#include "processes/DPW.hpp"
#include "processes/NGGP.hpp"
#include "processes/NGGPW.hpp"
#include "utils/Data.hpp"
#include "utils/Likelihood.hpp"
#include "utils/Params.hpp"
#include "utils/Process.hpp"
#include <chrono>

Functions
	RCPP_MODULE (params_module)
	Rcpp module to expose the Params class to R.
Rcpp::List	mcmc (const Eigen::MatrixXd &distances, Params &param, const Rcpp::IntegerVector &initial_allocations_r=Rcpp::IntegerVector())
	Main MCMC function for Bayesian non-parametric clustering.

Detailed Description

Main launcher file for MCMC simulation with various Bayesian non-parametric processes.

Author

Filippo Galli

Date

2025

This file contains the main MCMC implementation and R interface for running Bayesian non-parametric clustering using different processes (DP, DPW, NGGP, NGGPW) with various sampling strategies (Neal3, Split-Merge, Split-Merge SAMS).

Function Documentation

mcmc()

Rcpp::List mcmc	(	const Eigen::MatrixXd &	distances,
		Params &	param,
		const Rcpp::IntegerVector &	initial_allocations_r = Rcpp::IntegerVector() )

Main MCMC function for Bayesian non-parametric clustering.

This function performs Markov Chain Monte Carlo (MCMC) sampling for clustering analysis using various Bayesian non-parametric processes. It supports different process types (DP, DPW, NGGP, NGGPW) and sampling algorithms (Neal3, Split-Merge).

Parameters

distances	A matrix of distances between data points (Eigen::MatrixXd)
param	Reference to a Params object containing all MCMC parameters and hyperparameters
initial_allocations_r	Optional initial cluster allocations (Rcpp::IntegerVector). If empty, random initialization will be used.

Returns

Rcpp::List containing:

• allocations: List of cluster allocations for each iteration
• K: List of number of clusters for each iteration
• loglikelihood: Vector of log-likelihood values for each iteration
• U: Vector of process-specific parameter U for each iteration

The function:

1. Initializes data structures and likelihood computation
2. Sets up the chosen Bayesian non-parametric process (currently NGGPW)
3. Configures the sampling algorithm (currently Split-Merge)
4. Runs MCMC for burn-in + sampling iterations
5. Updates process parameters and saves results at each iteration
6. Provides progress output and timing information

Note

The process type and sampler can be changed by uncommenting/commenting the appropriate lines in the function body.

RCPP_MODULE()

RCPP_MODULE

(

params_module

)

Rcpp module to expose the Params class to R.

This module creates an R interface for the Params class, allowing R users to create and manipulate parameter objects for the MCMC simulation. The module exposes constructors and all parameter fields with their descriptions.