Skip to main content

Table 1 Comparison of different methods on their cluster validity in the simulation

From: Multi-view singular value decomposition for disease subtyping and genetic associations

 

e=1

e=0.8

e=0.6

e=0.4

Single-view SSVD

0.0821

0.1798

0.2432

0.2286

Co-regularized Spectral

0.2306

0.2477

0.2338

0.2549

Kernel addition

0.2587

0.2295

0.2350

0.2566

Kernel product

0.1917

0.2432

0.2302

0.2310

Feature concatenation

0.1569

0.1576

0.1532

0.1211

Proposed method

0.7949

0.7693

0.6815

0.6329

  1. The normalized mutual information (NMI) values are shown, measuring the agreement between the clusters resulting from an approach and the simulated phenotypic clusters. The genetic contribution to the phenotypic variation varied according to different e values. A greater e value indicates a higher agreement between the simulated phenotypic clusters and genotypic clusters, making it easier for a clustering approach to recover the simulated phenotypic clusters. Italic fonts indicate the best performance in the experiments with each of the e values.