PFA · Stochastic
Noisy channel: does it end in a?
A two-state probabilistic automaton over a 10% noisy channel. The accepting mass is 0.9 after an a and 0.1 after a b, so cut-point 0.5 recovers the language "ends in a" despite the noise.
The author’s examples, run
a → acceptb → rejectba → acceptab → rejectbbba → acceptaaab → reject