Label Distribution Learning on Natural (3 random splits)
0.5216KL DivergenceERDF
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ERDFnumber of trees=100, max depth=10, max cascade forest layers=10, early stopping tolerance=1, feature enhancement k=52026.02 | 0.5216 | 0.2221 | 2.3595 | 6.3066 | 0.8343 | 67.04 | |
| StructRFnumber of trees=100, max depth=102026.02 | 0.6689 | 0.2794 | 2.3465 | 6.2559 | 0.782 | 59.27 | |
| LDL-SCLhyperparameters=λ1, λ2, λ3 tuned from {10^-1, 10^-2, 10^-3}, number of clusters=5 or 102026.02 | 0.8283 | 0.3329 | 2.4682 | 6.7833 | 0.7284 | 49.94 | |
| AA-KNNk=5, distance metric=minkowski2026.02 | 1.0357 | 0.3134 | 1.9085 | 4.5686 | 0.715 | 56.21 | |
| AA-BPoptimizer=SA-BFGS, max iterations=5002026.02 | 1.2452 | 0.3922 | 2.5099 | 7.0532 | 0.5582 | 36.65 |