Correctness of Hybrid Guruswami-Sudan Interpolation #
The hybrid's fuel/fallback dispatch affects performance only: its result
always coincides with the result of one of the two verified backends
(hybridPositiveInterpolate_eq_lee_or_approximant). Soundness and
completeness of hybridInterpContext follow by case analysis from the
corresponding backend theorems.
The only reduction-level fact needed is determinism of the fueled
Mulders-Storjohann loop (muldersStorjohannReduceWithFuel_eq_of_no_conflict):
when the budgeted probe ends without a shifted leading conflict, it stopped at
the same matrix as the full reduction, so the kept Lee branch is literally the
Lee-O'Sullivan backend run with the fast reducer context.
The hybrid branch result always coincides with one of the two verified backends.
The hybrid result always coincides with one of the two verified backends.
Public hybrid interpolation backend context: adaptive Lee-O'Sullivan reduction under a step budget with an approximant-basis fallback.