Replay verdicts as verifiers #
Replaying a fixed fair-coin machine on an instance with a string of saved coins, and testing
whether the output is [true], is a verifier on encoded pairs (a, w). It runs in linear time and
linear space in the encoded pair, because replay itself does
(Turing.MultiTapePTM.computableInDegree_runClosedCoins) and the replayed output is no longer
than the replay time. Both readings of a time-bounded machine as a verifier use it: the
certificate as a computation path of a nondeterministic machine
(Cslib.Complexity.exists_verifier_of_mem_NTIME), and the random string of a probabilistic one
(Cslib.Complexity.exists_verifier_of_mem_PromiseBPTIME).
Main results #
Turing.MultiTapePTM.computableInO_decide_runClosedCoins: the replay verdict is computable in linear time and space in the encoded pair.Turing.MultiTapePTM.computableInPolyTime_decide_runClosedCoins: the replay verdict is polynomial-time.
Replay verdicts are linear-time verifiers. Replaying machine on enc a with the saved
coins w, and testing whether the output is [true], is computable in linear time and linear space
in the encoded pair (a, w).
Replay verdicts are polynomial-time verifiers.