Table 3 Evaluated algorithms and the range of parameters that have been tested.
From: Genome classification by gene distribution: An overlapping subspace clustering approach
| Â | Â | Test range | Â | ||
|---|---|---|---|---|---|
Algorithm | Parameter | Min. | Max. | Step size | Best case |
O-HARP | t – cluster tightness | 0.1 | 0.9 | 0.04 | 0.58 |
|  | f – filtering threshold | 0 | 0.04 | 0.8 | 0.24 |
HARP | K – target # of clusters | 3 | 30 | 1 | 7 |
|  | MOP – max. outlier percentage | 5 | 13 | 1 | 9 |
SAMBA | v – version | (discrete: 6 versions – tested all) | v2 | ||
|  | t – try covering all probes | (discrete: true/false – tested both) | true | ||
|  | f – overlap factor | 0.001 | 0.13 | varies | 0.03 |
|  | rp – responding probes to hash | 3 | 30 | 3 | 9 |
Cheng-Church | d – delta | 0.03 | 0.9 | varies | 0.07 |
|  | a – alpha | 1.0 | 1.2 | 0.1 | 1.2 |
|  | K – target # of clusters | 10 | 300 | varies | 70 |
FastDOC | K – target # of clusters | 3 | 18 | 3 | n/a |
|  | b – beta | 0.2 | 0.45 | 0.05 | n/a |
|  | w – cluster width | 0.05 | 0.65 | 0.2 | n/a |
|  | MAXITER – max. # of inner iterations | 8000 | 10000 | 2000 | n/a |
SynFPS | K – target # of clusters | 5 | 10 | 1 | 9 |
n-gram | n – length of nucleotide sequence | 2 | 6 | 1 | 5 |
|  | k – target # of clusters | 3 | 10 | 1 | 9 |