Dataset Colon 1

Basic characteristics Colon 1

22

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. Normal tissue is used as target class. Download mat-file with Prtools dataset.

40

outlier objects

1908

features

Unsupervised PCA Colon 1

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 1

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

Results Colon 1

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

570, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 70.4 (1.1) 71.3 (2.9) 48.8 (3.5)
Min.Cov.Determinant NaN (0.0) NaN (0.0) NaN (0.0)
Mixture of Gaussians NaN (0.0) NaN (0.0) 64.8 (6.3)
Naive Parzen 70.0 (1.5) 70.0 (1.5) 52.5 (5.5)
Parzen 36.4 (22.4) 36.4 (22.4) 66.4 (4.4)
k-means 66.8 (3.1) 71.6 (4.0) 55.9 (4.7)
1-Nearest Neighbors 71.3 (3.3) 74.3 (1.2) 66.6 (2.2)
k-Nearest Neighbors 71.3 (3.3) 74.3 (1.2) 66.6 (2.2)
Nearest-neighbor dist 72.1 (4.7) 70.3 (2.0) 67.2 (4.9)
Principal comp. 70.7 (1.6) 70.7 (1.9) 48.2 (4.1)
Self-Organ. Map 68.2 (2.6) 72.9 (1.9) 60.9 (8.1)
Auto-enc network NaN (0.0) NaN (0.0) 58.9 (4.3)
MST 69.3 (3.0) 73.5 (2.2) 65.2 (4.2)
L_1-ball 76.2 (3.6) 51.6 (5.2) 46.1 (3.4)
k-center 68.4 (2.9) 73.2 (1.8) 59.3 (1.0)
Support vector DD 36.4 (22.4) 36.4 (22.4) 36.4 (22.4)
Minimax Prob. DD 0.0 (0.0) 0.0 (0.0) 40.1 (5.7)
LinProg DD 41.8 (20.0) 36.8 (22.4) 66.8 (2.5)
Lof DD 67.2 (1.5) 73.3 (3.2) 35.5 (7.6)
Lof range DD NaN (0.0) NaN (0.0) NaN (0.0)
Loci DD 62.8 (2.5) 68.7 (3.5) 61.2 (2.3)

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