Algorithm 1.
Estimate optimal sigma and optimal area set
| 1: | Inputs: TrainV alSet |
| 2: | Outputs: SigmaOpt, Iopt |
| 3: | Define I(1): indexes for all areas |
| 4: | Define P: number of areas |
| 5: | for p = 1 to P − 1 do |
| 6: | Validate sigma, one RCK iteration (TrainV alSet, I(p)) ⇒ Sigma(p) and E(p) |
| 7: | Train with TrainV alSet, Sigma(p) and I(p) |
| 8: | Compute discriminative weights |
| 9: | Remove area with lowest weight |
| 10: | Store indexes of remaining areas ⇒ I (p + 1) |
| 11: | end for |
| 12: | Find p that minimizes E(p) ⇒ pmin |
| 13: | Sigma(pmin) ⇒ SigmaOpt, I (pmin) ⇒ I opt |