Benchmarks¶

Correctness fixes against published formulas and documented optima.

Moving Peaks `pf1` peak shape and `ALT1` severity¶

Branke scenario 1 defines \(P = h / (1 + w \cdot \sum_i (x_i - p_i)^2)\). A first pass restored Euclidean distance to match DEAP's wording; the published MPB scenario 1 is the squared form, and that is what MPFuncs.pf1 uses now. The DEFAULT docstring matches that form. MPConfigs.ALT1["move_severity"] is 1.5 (Branke scenario 2 / the preset table), not 1.0.

Validators.

tests/test_benchmarks/test_moving_peaks.py::test_pf1_squared_form_differs_from_euclidean
tests/test_benchmarks/test_moving_peaks.py::test_pf1_uses_euclidean_distance_and_alt1_move_severity

DTLZ5 / DTLZ6 \(f_1\) multiplies extra cosines¶

_dtlz_helper_2 took \(\prod \cos\theta\) over individual[1:], so distance variables that already feed \(g\) were folded into \(f_1\) again. \(\sum f_i^2 = (1+g)^2\) broke whenever len(individual) > count.

Fix. The \(f_1\) product uses individual[1:count-1] only — the same angular slice as \(f_2 \ldots f_M\).

Validator. tests/test_benchmarks/test_multi_obj.py::test_dtlz5_f1_uses_only_the_angular_variables

Chuang F3 all-ones scored 31, not 40¶

The selector-1 branch kept _inv_trap on the rotated blocks, stopped one block early (bit 38 unscored), and put the selector bit inside the wrap trap. All-zeros scored 40; all-ones scored 31.

Fix. Selector-1 blocks use _trap. The 2-bit rotation covers bits 2:38 in steps of 4 plus wrap [38, 39, 0, 1]. Bit 40 is selector only. All-ones and all-zeros both score 40.

Validator. tests/test_benchmarks/test_binary.py::test_chuang_f3_optima_and_rotated_trap_blocks

Kotanchek used 3.2 instead of 1.2¶

The TEVC 2009 / GPTP target is \(1 / (1.2 + (x_2-2.5)^2)\). The implementation used 3.2.

Fix. Both the function and the docstring table use 1.2.

Validator. tests/test_benchmarks/test_symb_regr.py::test_kotanchek_uses_published_1_2_denominator

Royal Road R2 dropped the top-level schema¶

The hierarchy walk stopped before n_order == len(individual), so the full-length schema was never scored. Classic R2 (64 bits, order 8) could not reach 256.

Fix. Walk n_order <= len(individual) so the top schema is included.

Validator. tests/test_benchmarks/test_binary.py::test_royal_road_2_classic_optimum_includes_top_schema

Royal Road decoded `True`/`False` as `"True"`/`"False"`¶

int("".join(map(str, values)), 2) raised ValueError: invalid literal ... 'TrueTrue'. mut_flip_bit preserves bool, so a boolean individual crashed on evaluate.

Fix. Coerce each bit with int(bool(v)) before joining, the same decode as bin2float.

Validator. tests/test_benchmarks/test_binary.py::test_royal_road_decodes_boolean_bits_like_integers