Table 2 |
||||||
|
Results for different training set sizes. |
||||||
|
Classification Accuracy (%) |
||||||
|
|
||||||
|
No. of training data |
Similar Classes |
Dissimilar Classes |
All Classes |
|||
|
|
||||||
|
No updating |
With updating |
No updating |
With updating |
No updating |
With updating |
|
|
|
||||||
|
50 |
82.2 |
91.3 |
95.3 |
98.5 |
90.1 |
95.6 |
|
40 |
80.8 |
90.2 |
94.9 |
98.3 |
89.2 |
95.1 |
|
30 |
78.9 |
88.9 |
94.2 |
98.4 |
88.1 |
94.6 |
|
20 |
76.3 |
87.5 |
93.2 |
98.0 |
86.4 |
93.8 |
|
10 |
71.2 |
80.8 |
90.6 |
96.6 |
82.9 |
90.3 |
|
|
||||||
|
Base SVM classifiers were trained using various numbers of cells for each of ten classes. Graphical models using those base classifiers were then tested on fields containing six cells each from two classes and results were averaged over 10 trials each for all possible pairs. |
||||||
|
Chen and Murphy BMC Bioinformatics 2006 7:90 doi:10.1186/1471-2105-7-90 |
||||||
