Edited nearest neighbour classifier. Instance selection. MATLAB code.
- functions
[n, stored, e] = edit_greedy_tabu_search(Data, Labels, tabu_gap, verbose)
[n,e] = edit_closest_to_centroid(Data,Labels)
C = train_1nn(TrainingData, TrainingLabels,~)
[e, AssignedLabels] = test_1nn(C,Data,Labels)
f = voronoi_regions(prototypes, region2d, colour)
- scripts
Example_GTS
This code applies the greedy tabu search method (GTS) to extract one prototype per class. It plots two figures: (1) The starting point of the GTS, which is the closest-to-centroid selection of prototypes (CC), and (2) The end of the GTS algorithm with the re-positioned prototypes. The Voronoi cells defined by the prototypes are shaded in a pastel version of the colour of the class.
- data
Data2D5.mat data file needed for the example in Example_GTS
- figures
Data2D5_CC.jpg and Data2D5_GTS.jpg - output from script Example_GTS.