x5tne/pdbs-v1-results
i benchmark my es, Pivot-Directional Basin Search, and i think it's better (dim: 10) Objective old pdbs pdbs-linear pdbs-quadratic PDBS-v1 sphere (mean) 3.2e-9 0.066 0.0025 2.6e-11 rosenbrock (mean) 0.36 22.8 94.1 (unstable) 0.50 rastrigin (mean) 40.0 (stuck) 40.0 (stuck) 1.87 16.1 ackley (mean) 1.76 6.59 (stuck) 0.073 6.8e-6 griewank (mean) 0.0070 0.0142 0.0048 0.0059 (dim:10, 20, 50) Objective Dim CMA-ES (mean fitness) CMA-ES time (s) PDBS-v1 (mean fitness)… See the full description on the dataset page: https://huggingface.co/datasets/x5tne/pdbs-v1-results.
i benchmark my es, Pivot-Directional Basin Search, and i think it's better
(dim: 10) | Objective | old pdbs | pdbs-linear | pdbs-quadratic | PDBS-v1 | |---|---|---|---|---| | sphere (mean) | 3.2e-9 | 0.066 | 0.0025 | 2.6e-11 | | rosenbrock (mean) | 0.36 | 22.8 | 94.1 (unstable) | 0.50 | | rastrigin (mean) | 40.0 (stuck) | 40.0 (stuck) | 1.87 | 16.1 | | ackley (mean) | 1.76 | 6.59 (stuck) | 0.073 | 6.8e-6 | | griewank (mean) | 0.0070 | 0.0142 | 0.0048 | 0.0059 |
(dim:10, 20, 50) | Objective | Dim | CMA-ES (mean fitness) | CMA-ES time (s) | PDBS-v1 (mean fitness) | PDBS-v1 time (s) | |---|---|---|---|---|---| | sphere | 10 | 7.3×10⁻⁴⁹ | 1.7 | 2.6×10⁻¹¹ | 1.9 | | sphere | 20 | 4.5×10⁻⁴⁵ | 7.9 | 0.00136 | 2.2 | | sphere | 50 | 2.2×10⁻²⁹ | 23.7 | 0.289 | 4.1 | | rosenbrock | 10 | 0.399 | 1.9 | 0.500 | 2.2 | | rosenbrock | 20 | 0.199 | 10.1 | 4.88 | 2.0 | | rosenbrock | 50 | 2.26 | 23.5 | 13.5 | 4.1 | | rastrigin | 10 | 13.5 | 2.8 | 16.1 | 2.3 | | rastrigin | 20 | 32.9 | 10.5 | 80.0 | 1.8 | | rastrigin | 50 | 98.8 | 23.9 | 200.0 | 3.8 | | ackley | 10 | 4.0×10⁻¹⁵ | 2.9 | 6.8×10⁻⁶ | 2.4 | | ackley | 20 | 4.0×10⁻¹⁵ | 10.3 | 6.57* | 1.9 | | ackley | 50 | 9.5×10⁻¹⁵ | 23.9 | 6.59* | 3.9 | | griewank | 10 | 0.0572 | 2.8 | 0.0059 | 2.2 | | griewank | 20 | 0.0304 | 10.3 | 0.0061** | 2.3 | | griewank | 50 | 0.00037 | 23.1 | 0.0098 | 4.4 |
*all 20 PDBS-v1 seeds land on the exact same value