Table 4.
Error rate in scoring TMA images. Note that the first row corresponds to results obtained by RF on the original set of features (i.e., without deep features).
| Deep features | # clusters or leaf nodes |
Error rate |
|---|---|---|
| — | — | 24.79% |
| K-means | 40 | 24.02% |
| Diana | [10,40] | 24.20% |
| Agenes | [10,40] | 24.14% |
| hclust | [10,40] | 24.29% |
| Agenes + Diana | [10,40] | 23.77% |
| Agenes + hclust | [10,40] | 23.71% |
| hclust + Diana | [10,40] | 23.52% |
| Agenes + Diana + hclust | [10,40] | 23.46% |
| rpTrees | 30 | 23.28% |