Edytuj stronę Odnośniki Fold/unfold all ODT export Ta strona jest tylko do odczytu. Możesz wyświetlić źródła tej strony ale nie możesz ich zmienić. ====== Constrained LVQ - Porownanie klasyfikatorów - Test 10CV ====== ^ Dataset ^^^^ QPC/LVQ1 ^^^^^^ PCA/LVQ1 ^^^^^^ LVQ1 ^^^ SVM ^^^^ kNN ^^^^ MLP ^^^ | | vec | feat | cl | acc. | std. | #P | std. | #K | std. | acc. | std. | #P | std. | #K | std. | acc. | std. | #K | acc. | std. | #SV | std. | acc. | std. | #K | std. | acc. | std. | #N | | Appendicities | 106 | 7 | 2 | **86.1** | 8.5 | 1.0 | 0.0 | 2.5 | 1.2 | **82.2** | 7.4 | 1.0 | 0.0 | 3.1 | 1.4 | **86.6** | 6.8 | 2 | **86.7** | 10.8 | 31.4 | 2.9 | **84.0** | 6.1 | 4.8 | 1.7 | **87.6** | 9.3 | 0 | | Australian | 690 | 14 | 2 | **86.1** | 4.1 | 1.0 | 0.0 | 2.0 | 0.0 | **86.1** | 4.1 | 1.0 | 0.0 | 2.0 | 0.0 | **85.6** | 4.6 | 2 | **84.9** | 1.6 | 206.1 | 4.6 | **85.1** | 2.9 | 8.2 | 1.9 | **87.0** | 5.6 | 1 | | BitSymmetry 2 | 1000 | 20 | 2 | **87.3** | 4.2 | 1.2 | 0.4 | 3.5 | 1.0 | **78.1** | 11.4 | 1.8 | 0.7 | 5.3 | 1.7 | **75.7** | 4.3 | 12 | **98.0** | 1.4 | 296.5 | 7.8 | **89.3** | 2.6 | 5.8 | 6.2 | **96.7** | 2.4 | 2 | | BitSymetry15 | 100 | 15 | 2 | **82.0** | 11.7 | 1.0 | 0.0 | 3.0 | 0.0 | **53.0** | 14.9 | 2.0 | 0.8 | 4.3 | 2.3 | **68.0** | 12.5 | 18 | **75.0** | 13.5 | 80.0 | 8.4 | **72.0** | 14.7 | 2.5 | 2.7 | **85.0** | 10.8 | 2 | | BitSymetry20 | 100 | 20 | 2 | **81.0** | 7.0 | 1.0 | 0.0 | 3.0 | 0.0 | **58.0** | 9.8 | 1.9 | 0.8 | 4.2 | 1.3 | **59.0** | 15.8 | 8 | **70.0** | 14.2 | 84.7 | 6.9 | **68.0** | 11.3 | 2.9 | 2.8 | **76.0** | 15.1 | 2 | | Breast Cancer W. | 683 | 9 | 2 | **96.2** | 1.8 | 1.0 | 0.0 | 2.0 | 0.0 | **96.2** | 2.1 | 1.0 | 0.0 | 2.0 | 0.0 | **96.5** | 2.5 | 2 | **96.6** | 1.7 | 51.1 | 3.1 | **97.1** | 1.4 | 5.6 | 2.1 | **96.6** | 1.4 | 0 | | Czerniak (trs) | 250 | 14 | 4 | **85.7** | 6.2 | 1.0 | 0.0 | 4.0 | 0.0 | **70.8** | 9.0 | 1.7 | 0.6 | 4.3 | 0.6 | **76.8** | 7.2 | 4 | **85.2** | 5.7 | 240.3 | 23.0 | **86.0** | 7.4 | 1.0 | 0.0 | **96.0** | 2.7 | 4 | | Glass | 214 | 9 | 6 | **60.4** | 10.3 | 1.1 | 0.3 | 4.5 | 0.9 | **58.9** | 7.3 | 1.4 | 0.6 | 4.0 | 0.5 | **66.3** | 10.5 | 7 | **64.9** | 6.2 | 283.6 | 17.6 | **68.8** | 9.2 | 1.4 | 0.8 | **69.2** | 10.4 | 7 | | Heart | 270 | 13 | 2 | **78.9** | 8.5 | 1.0 | 0.0 | 2.0 | 0.0 | **80.7** | 9.0 | 1.0 | 0.0 | 2.0 | 0.0 | **82.2** | 8.6 | 2 | **81.5** | 9.1 | 101.5 | 7.0 | **78.5** | 7.6 | 8.5 | 1.2 | **82.6** | 6.8 | 0 | | Ionosphere | 200 | 34 | 2 | **78.4** | 8.0 | 1.0 | 0.0 | 3.1 | 0.7 | **75.5** | 8.5 | 1.1 | 0.3 | 3.5 | 0.8 | **81.9** | 7.0 | 4 | **93.5** | 4.7 | 61.0 | 4.1 | **84.0** | 7.7 | 1.2 | 0.6 | **89.4** | 8.1 | 3 | | Iris | 150 | 4 | 3 | **96.0** | 4.4 | 1.0 | 0.0 | 3.0 | 0.0 | **94.7** | 4.0 | 1.0 | 0.0 | 3.0 | 0.0 | **97.3** | 3.3 | 3 | **96.7** | 4.7 | 39.6 | 5.5 | **94.5** | 6.9 | 5.8 | 3.1 | **94.7** | 8.2 | 0 | | L. Breast | 277 | 9 | 2 | **72.6** | 4.4 | 1.0 | 0.0 | 2.0 | 0.0 | **74.0** | 6.9 | 1.0 | 0.0 | 2.1 | 0.3 | **74.7** | 6.0 | 3 | **73.3** | 9.6 | 143.6 | 4.5 | **73.7** | 5.5 | 6.9 | 2.7 | **75.8** | 5.6 | 2 | | Led500 | 500 | 7 | 10 | **58.5** | 8.0 | 1.1 | 0.3 | 9.6 | 0.5 | **45.5** | 8.6 | 1.5 | 0.7 | 8.2 | 1.1 | **72.0** | 5.1 | 10 | **65.2** | 5.5 | 664.1 | 18.1 | **71.2** | 6.6 | 8.5 | 0.8 | **70.8** | 7.5 | 0 | | Parity 10 | 1024 | 10 | 2 | **96.1** | 3.9 | 1.0 | 0.0 | 6.8 | 0.4 | **97.6** | 0.8 | 1.0 | 0.0 | 7.5 | 0.7 | **51.3** | 4.9 | 8 | **44.2** | 5.7 | 921.2 | 1.0 | **80.7** | 3.4 | 20.0 | 0.0 | **97.6** | 1.3 | 8 | | Parity 8 | 256 | 8 | 2 | **94.5** | 5.0 | 1.0 | 0.0 | 6.3 | 0.6 | **96.9** | 3.8 | 1.0 | 0.0 | 6.7 | 0.5 | **52.0** | 8.0 | 9 | **35.4** | 11.9 | 119.3 | 1.7 | **100.0** | 17.0 | 0.0 | 0.0 | **96.1** | 4.8 | 10 | | Voting | 435 | 16 | 2 | **95.2** | 3.5 | 1.0 | 0.0 | 2.0 | 0.0 | **90.6** | 5.5 | 1.3 | 0.5 | 2.5 | 0.9 | **93.8** | 2.7 | 5 | **95.9** | 2.4 | 57.0 | 10.6 | **93.3** | 3.2 | 4.6 | 2.5 | **94.0** | 3.5 | 0 | | Wine | 178 | 13 | 3 | **96.0** | 5.7 | 1.9 | 0.3 | 3.1 | 0.3 | **94.9** | 4.7 | 2.0 | 0.0 | 3.0 | 0.0 | **97.7** | 2.8 | 4 | **96.6** | 2.9 | 63.7 | 8.0 | **95.0** | 4.1 | 6.2 | 3.5 | **98.3** | 2.7 | 0 | Breast Cancer Wisconsin and L. Breast data without vectors with missing values #P number of projections (hidden nodes) #K number of prototypes (output nodes) or number of nearest neighbours (kNN) #SV number of support vectors #N number of hidden nodes (MLP) [[projects:models:llvq:cvtest:mlp2|MLP Szczegółowe wyniki]] \\ [[projects:models:llvq:cvtest:lvq_all|LVQ Szczegółowe wyniki]] \\ [[projects:models:llvq:cvtest:mlp_matlab_all|MLP Matlab Toolbox]] \\ [[projects:models:llvq:cvtest:mlp_snns_par|SNNS na binarnych]] \\ [[projects:models:llvq:cvtest:mlpporownanie|MLP porownanie rozbierznosci wynikow]] \\ [[projects:models:llvq:cvtest:mlp|MLP Szczegółowe wyniki (z kara)]] \\ ===== Settings ===== ==== Test ==== 10 CV, stratified Data transformation: normalization ==== QPC/LVQ ==== Network parameters (Matlab) Method lvq+qpc Initialization 5 LVQ learnign rate 0.010000 Attraction force 0.050000 Max. projections 5 K range 2 10 20 Attraction step 100 Precision (eps) 0.030000 QPC parameters (default) lrate = 0.1; % learning rate maxiter = 1000; % MAX number of iterations init = 5; % number of initializations function = gauss; % width = 0.1; % function width eps = 0.001; % precision ==== PCA/LVQ ==== Network parameters (Matlab) Method lvq+pca Initialization 5 LVQ learnign rate 0.010000 Attraction force 0.050000 Max. projections 5 K range 2 10 20 Attraction step 10 Precision (eps) 0.030000 ==== LVQ1 ==== Initialization 5 LVQ learnign rate 0.010000 K range 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Precision (eps) 0.010000 ==== SVM ==== Ghost Miner 3.0 beta5 (21-05-2004) Gaussian kernel Auto C & Bias (5CV, stratified) ==== MLP ==== C++ implementation transfer function : sigmoid optimization : back propagation initiations : 10 hidden nodes : from 0 to 20 ==== kNN ==== Ghost Miner 3.0 beta5 (21-05-2004) Auto K (5CV, stratified) Euclidean distance measure Max K=10