Dataset Leukemia 1

Basic characteristics Leukemia 1

25

target objects

The Leukemia gene expression dataset, presented in T.R.Golub et al., Molecular Classification of Cancer: class discovery and class prediction by gene expression monitoring, Science, 1999. Download mat-file with Prtools dataset.

47

outlier objects

3571

features

Unsupervised PCA Leukemia 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 Leukemia 1

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

Results Leukemia 1

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

572, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 94.7 (2.7) 94.5 (2.9) 65.3 (3.0)
Min.Cov.Determinant NaN (0.0) NaN (0.0) NaN (0.0)
Mixture of Gaussians NaN (0.0) NaN (0.0) NaN (0.0)
Naive Parzen 96.7 (0.4) 96.7 (0.4) 62.7 (5.4)
Parzen NaN (0.0) 41.1 (30.2) 67.6 (1.6)
k-means 89.2 (2.4) 94.7 (2.8) 58.7 (7.0)
1-Nearest Neighbors 92.5 (2.3) 94.9 (2.6) 71.4 (6.2)
k-Nearest Neighbors 92.5 (2.3) 94.9 (2.6) 71.4 (6.2)
Nearest-neighbor dist 87.3 (2.9) 93.6 (2.5) 57.6 (4.8)
Principal comp. 94.5 (2.5) 94.5 (2.9) 58.5 (4.7)
Self-Organ. Map 88.3 (1.3) 94.7 (2.7) 58.2 (1.1)
Auto-enc network NaN (0.0) NaN (0.0) 58.1 (3.5)
MST 88.6 (0.9) 94.7 (2.8) 71.1 (3.1)
L_1-ball 42.4 (29.3) 42.4 (29.3) 56.5 (6.3)
k-center 92.2 (2.6) 95.3 (2.5) 67.4 (5.6)
Support vector DD 41.1 (30.2) 41.1 (30.2) 41.1 (30.2)
Minimax Prob. DD 92.5 (2.3) 0.0 (0.0) 71.4 (6.2)
LinProg DD 93.3 (2.2) 41.1 (30.2) 71.7 (2.7)
Lof DD 90.0 (2.8) 93.8 (1.9) 53.6 (2.8)
Lof range DD 90.7 (2.4) 94.4 (2.6) 57.9 (1.8)
Loci DD 84.8 (3.1) 91.8 (3.1) 49.0 (0.9)

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