Dataset Colon 2

Basic characteristics Colon 2

40

target objects

The Colon gene expression dataset, presented in U.Alon et al., Broad patterns of gene expression revealed by clustering of tumor and normal colon tissues probed by oligonucleotide arrays, PNAS, 1999. Tumor tissue is used as target class. Download mat-file with Prtools dataset.

22

outlier objects

1908

features

Unsupervised PCA Colon 2

On the left, the PCA scatterplot is shown, on the right the retained variance for varying number of features.
On the left, the PCA scatterplot is shown of data rescaled to unit variance, on the right the retained variance.

Supervised Fisher Colon 2

On the left, the Fisher scatterplot is shown, on the right the ROC curve along this direction.

Results Colon 2

The experiments are performed using dd_tools. A rudimentary explanation of the classifiers is given in the classifier section.

571, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 61.1 (3.8) 68.4 (3.6) 72.1 (2.9)
Min.Cov.Determinant NaN (0.0) NaN (0.0) NaN (0.0)
Mixture of Gaussians NaN (0.0) NaN (0.0) 66.2 (5.7)
Naive Parzen 73.4 (3.1) 73.7 (3.2) 71.0 (3.1)
Parzen 63.6 (22.4) 63.6 (22.4) 69.1 (3.2)
k-means 67.6 (2.5) 72.3 (1.9) 67.6 (5.7)
1-Nearest Neighbors 64.4 (3.4) 70.8 (2.8) 66.4 (3.4)
k-Nearest Neighbors 64.4 (3.4) 70.8 (2.8) 66.4 (3.4)
Nearest-neighbor dist 43.9 (5.9) 53.3 (7.4) 40.4 (7.1)
Principal comp. 60.3 (2.9) 68.4 (3.5) 46.0 (3.4)
Self-Organ. Map 68.3 (1.4) 71.9 (4.0) 64.8 (4.2)
Auto-enc network NaN (0.0) NaN (0.0) 66.5 (4.7)
MST 65.2 (5.3) 72.2 (2.2) 65.6 (3.2)
L_1-ball 72.2 (3.7) 57.1 (3.6) 55.4 (6.5)
k-center 61.3 (6.4) 68.1 (2.1) 61.8 (6.4)
Support vector DD 63.6 (22.4) 63.6 (22.4) 63.6 (22.4)
Minimax Prob. DD 0.0 (0.0) 0.0 (0.0) 29.8 (2.2)
LinProg DD 67.6 (19.4) 67.4 (19.4) 66.2 (3.5)
Lof DD 59.8 (4.4) 66.4 (2.5) 47.2 (3.8)
Lof range DD 62.3 (1.5) 71.9 (3.5) 47.0 (5.5)
Loci DD 66.5 (1.5) 71.8 (2.3) 58.8 (2.9)

Classifier projection spaces The first classifier projection spaces are obtained by computing the classifier label disagreements (setting the threshold on 10% target error) and applying an MDS on the resulting distance matrix between classifiers:



Original



Unit variance



PCA mapped

Classifier projection spaces The second versions of the classifier projection spaces are obtained by computing the classifier ranking disagreements and applying an MDS on the resulting distance matrix between classifiers:



Original



Unit variance



PCA mapped