adanjoga / cvik-toolbox

CVIK is a Toolbox for the automatic determination of the number of clusters on data clustering problems

Adan Jose-Garcia / Homepage / Twitter / Linkedin

I am a Research Scientist in Machine Learning and Digital Health at the CRIStAL and INFINITE Labs, University of Lille, France. My research focuses on developing integrative cluster analysis approaches to address healthcare-related data problems and help to understand better disease complications and treatment goals.

Research projects

• BicViz A tool to visualize biclusters.
• CVIk Toolbox A toolbox of cluster validity indices for determining the number of clusters on clustering problems.
• C3-IoC A career guidance system for assessing student skills.
• EvoClustering A repository of publications, datasets, and source codes for evolutionary clustering algorithms.

Selected publications

• Metaheuristic Biclustering Algorithms: From State-of-the-art to Future Opportunities (2023)
• Evolutionary Multi-objective Clustering Over Multiple Conflicting Data Views (2022)
• An Evolutionary Many-objective Approach to Multiview Clustering Using Feature and Relational Data (2021)
• Automatic Clustering Using Nature-inspired Metaheuristics: A Survey (2016)

cvik-toolbox's Issues

Alternate Pairwsie and Information Based Indices

There are other algorithms out there for external validation, as noted in GiulioRossetti/cdlib#147 (comment)

Error with own distance function when calculating Davies Bouldin

I wanted to use my own distance function, as below:

distfun = @(XI,XJ)(1-abs(MyCorr(XI',XJ')));

MyCorr function:
function c = MyCorr(in1,in2)

% Average reference both sets of input maps
nO = size(in1,1);
Centering = eye(nO) - 1/nO;

in1 = Centering * in1;

if nargin < 2
    in2 = in1;
else
    in2 = Centering * in2;
end

% Compute covariance matrix between sets of input maps
covar = in1' * in2;
var1 = diag(in1' * in1);
var2 = diag(in2' * in2);

c = covar./sqrt(var1*var2');

While it works with the silhouette coefficient function, I get following error with Davies Bouldin:

Error using internal.stats.parseArgs
Invalid parameter name: Distance.

Error in clustering.evaluation.DaviesBouldinEvaluation (line 53)
      internal.stats.parseArgs(basePnames, baseDflts, varargin{:});

Error in evalclusters (line 190)
           obj = clustering.evaluation.DaviesBouldinEvaluation(X,YorFun,varargin{:});

Does someone no the whats the issue?
Thanks a lot in advance!