benchmarks/math/README.md

Math Benchmarks

Mathematical optimization and algorithm evolution problems.

Problems

Signal processing & geometry (from SkyDiscover demos)

• signal_processing — Real-time adaptive filtering for non-stationary time series
• circle_packing — Pack 26 circles in a unit square to maximize sum of radii (AlphaEvolve B.12)

AlphaEvolve mathematical problems

12 problems from AlphaEvolve Appendices A and B. All evaluators are normalized to maximize the target metric.

Appendix A:

• matmul — Faster algorithm for matrix multiplication (A)

Appendix B:

1. first_autocorr_ineq — Upper bound on autoconvolution constant (B.1)
2. second_autocorr_ineq — Lower bound on autoconvolution norm constant (B.2)
3. third_autocorr_ineq — Upper bound on absolute autoconvolution constant (B.3)
4. uncertainty_ineq — Upper bound on Fourier uncertainty constant (B.4)
5. erdos_min_overlap — Upper bound on Erdos minimum overlap constant (B.5)
6. sums_diffs_finite_sets — Lower bound on sums/differences of finite sets (B.6)
7. hexagon_packing — Pack unit hexagons in a regular hexagon, n=11,12 (B.7)
8. minimizing_max_min_dist — Minimize max/min distance ratio, n=16 d=2 and n=14 d=3 (B.8)
9. heilbronn_triangle — Heilbronn problem for triangles, n=11 (B.9)
10. heilbronn_convex — Heilbronn problem for convex regions, n=13,14 (B.10)
11. circle_packing_rect — Pack circles in a rectangle of perimeter 4 (B.13)

Run

uv run skydiscover-run \
  benchmarks/math/signal_processing/initial_program.py \
  benchmarks/math/signal_processing/evaluator.py \
  -c benchmarks/math/signal_processing/config.yaml \
  -s [your_algorithm] \
  -i 100

Each problem directory contains initial_program.py, evaluator.py, and either config.yaml or per-search configs. Some multi-variant problems have numbered subdirectories (e.g., heilbronn_convex/13/, hexagon_packing/11/).