Exact

Exact Union RegExp

Repository files (latest commit first)

Filename	Latest commit message	Latest commit date
Transaction Science 86c2a5c393 qec_memory: held-out shot windows, every decoder's predictions kept, and paired tests against the published decoders ... The surface-code memory example (examples/qec_memory) gains what a paired comparison with the published decoders needs: - survey --skip K decodes shots K to K + N of each experiment (K even, so every shot keeps its parity and the model that parity picks), and scores the published decoders on the same shots: shots a decoder's settings were never chosen on. - survey --predictions DIR now also writes the published decoders' predictions on the same shots and the record's own flips, one byte a shot, beside this crate's. - Every receipt records its first shot, and per experiment, for each of this crate's decoders against each published decoder, the shots only one of the two got wrong. - qec_memory pairs SET KEY@RECEIPT... tests those counts: the exact two-sided McNemar test per distance and over every distance given, the fitted rates and Lambda of both, and a paired bootstrap (every fitted experiment's shots drawn again with the two decoders' outcomes on a shot kept together, each replicate from its own seed) of each distance's rate ratio and of the difference in Lambda. No receipt changes here; the runs that use it follow. Co-Authored-By: Claude Opus 5.5 <[email protected]>		2026-11-13 07:31:16 -05:00
..
data.rs	qec_memory: held-out shot windows, every decoder's predictions kept, and paired tests against the published decoders	2026-11-13 07:31:16 -05:00
main.rs	qec_memory: held-out shot windows, every decoder's predictions kept, and paired tests against the published decoders	2026-11-13 07:31:16 -05:00
pairs.rs	qec_memory: held-out shot windows, every decoder's predictions kept, and paired tests against the published decoders	2026-11-13 07:31:16 -05:00
receipt.rs	qec_memory: CPU time a shot in the rows (local)	2026-11-12 11:07:06 -05:00