Skip to content

Parity games: tangle learning, strategies, letter game (WP7) - #155

Merged
leonbohn merged 2 commits into
mainfrom
games-tangle-learning
Nov 9, 2026
Merged

leonbohn merged 2 commits into
mainfrom
games-tangle-learning

Conversation

@leonbohn

@leonbohn leonbohn commented Nov 9, 2026

Copy link
Copy Markdown
Owner

Remaining WP7 items of #143.

  • ParityGame::solve now runs tangle learning (as in Oink, ~300 lines on the existing attractor code) and returns winning strategies (Solution::strategy). Zielonka stays as solve_zielonka (reference; it returns strategies too).
  • ParityGame::letter_game: arena of a (non)deterministic parity automaton, the chooser picks letters, the opponent resolves nondeterminism.
  • Tests: proptest comparing both solvers and checking that the strategies win (up to 250 vertices, 30000 cases locally, the Zielonka fallback for a stuck search was never hit); letter game test (universality). Benchmarks: tangle_learning* next to zielonka*.
  • Instruction counts (gungraun, new vs. old): tangle learning is ~2x faster on ladder_4000 and many_priorities_4000, but ~2x slower on few_priorities (small random games, where Zielonka is already cheap). Zielonka got 6-22% slower from tracking strategies.

Not done: Python exposure of strategies / letter_game; ticking the boxes in #143.

@leonbohn leonbohn added the performance Runtime or memory performance label Nov 9, 2026
@leonbohn
leonbohn merged commit 67ebd2a into main Nov 9, 2026
15 checks passed
@leonbohn
leonbohn deleted the games-tangle-learning branch November 9, 2026 21:11
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

performance Runtime or memory performance

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant