Potential fixes for #48 (assisted by copilot)

[FMKerckhof]

Hi @mlampros - I got stuck in trying to resolve #48 so I prompted GitHub Copilot to come up with a fix for the predict function. I tested the proposed solution, and rebuilt the documentation. On the vignette example the code works smoothly, and I will validate this on a larger dataset as well.

The following code works:

library(ClusterR)

data("dietary_survey_IBS")

X <- center_scale(dietary_survey_IBS[, -ncol(dietary_survey_IBS)])

gmm <- GMM(X, gaussian_comps = 3, 
           full_covariance_matrices = TRUE,
           dist_mode = "eucl_dist",
           seed_mode = "random_subset", verbose=TRUE)
           
predictions <- predict_GMM(X, gmm$centroids, 
                           gmm$covariance_matrices, 
                           gmm$weights)

I have created this as a draft PR for now, until I can validate on a larger dataset

[Copilot]

Pull Request Overview

This PR extends the GMM prediction functionality to support full covariance matrices, addressing issue #48. The implementation adds a new C++ function to handle 3D covariance arrays while maintaining backward compatibility with diagonal covariance matrices.

Key changes:

• Added new C++ function predict_MGausDPDF_full for handling full covariance matrices (3D arrays/cubes)
• Enhanced predict_GMM function to automatically detect and handle both diagonal and full covariance matrices
• Added comprehensive test coverage for the new functionality

Reviewed Changes

Copilot reviewed 10 out of 11 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
inst/include/ClusterRHeader.h	Implements core C++ logic for full covariance matrix prediction
src/export_inst_folder_headers.cpp	Exports the new C++ function to R
R/clustering_functions.R	Updates predict_GMM to handle both diagonal and full covariance matrices
tests/testthat/test-gmm.R	Adds comprehensive test coverage for full covariance functionality
src/RcppExports.cpp	Auto-generated Rcpp exports for new function
src/init.c	Registers new C function
R/RcppExports.R	Auto-generated R wrapper for new function
man/predict_GMM.Rd	Updates documentation to reflect new functionality
DESCRIPTION	Version bump for RoxygenNote
.Rbuildignore	Adds R project files to ignore list

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[mlampros]

hi @FMKerckhof and thank you for the pull request. I took a look to the following files (mainly):

• test-gmm.R
• clustering_functions.R
• ClusterRHeader.h

The important changes are in the ClusterRHeader.h inside the

for (unsigned int i = 0; i < WEIGHTS.n_elem; i++) {} loop where both the WEIGHTS vector and the COVARIANCE slices have the same length. Thus, for each i a different slice (matrix) is computed compared to what it happens when full_covariance_matrices is FALSE, where each time the same matrix is used in the computation.

We can proceed and merge your pull request but could you adjust the DESCRIPTION file and add your data as a "Github Contributor"? ( person(given = "", family = "", email = "@email.com", role = "ctb", comment = "Github Contributor"))

I also found a small bug in the normalization constant where the determinant is used, which I'll fix once this pull request is merged.

[FMKerckhof]

@mlampros - done, feel free to merge 👍

[mlampros]

@FMKerckhof thank you.