Sequential mapping algorithm in MATLAB for performing cross-situational learning and mapping language to sensoric modality (or in general any two clusterings together) as used for a paper submitted to conference KUZ 2017
The function finds the best mapping between visual clustering of DataVis and language clustering LangLabs where both clusterings are reclustered after each clusters mapping
Inputs(1).Data = Visual_data; %now clustered by GMM
Inputs(2).Data = Language_data;
labelsNmb = True_labels; %vector of numerical values indicating class for each datapoint (just for evaluation purposes)
Pars.nmbClusters=[9]%number of clusters in visual data
Pars.init=1%1-random,2-Ikm
Pars.regularize = [1];%value by which regularize in gmdistribution while fitting Inputs(i).Data
Pars.smax = 5;%how many iterations of sequential mapping should be used
Results = sequential_mappingKUZ(Inputs,labelsNmb,Pars)