Theorem 5.7 (i) — the limit set of a precompact asymptotic pseudotrajectory is internally chain transitive
ProvedStochApproxDyn.LimitSet.limitSet_isInternallyChainTransitiveLet be a semiflow on an arbitrary metric space (not assumed compact, complete or locally compact). Let be an asymptotic pseudotrajectory of which is precompact, i.e. its image has compact closure in . Then the limit set
is internally chain transitive: it is nonempty, compact and invariant ( for all ), and for all and all , there is a -pseudo-orbit of from to all of whose points lie in .
This is the limit set theorem of Benaïm and Hirsch. Combined with the fact that interpolated stochastic approximation processes are asymptotic pseudotrajectories of the mean ODE, it says that such algorithms converge to internally chain transitive sets of the ODE, and so reduces their asymptotic analysis to the dynamics of the ODE.
Formalization Note A semiflow is Mathlib's Flow ℝ≥0 M and is a function ℝ≥0 → M; continuity of is part of being an asymptotic pseudotrajectory. Nonemptiness of is part of the conclusion, not a hypothesis.
import Mathlib import Definitions.Def_StochApproxDyn_LimitSet_AsymptoticPseudotrajectory import Definitions.Def_StochApproxDyn_LimitSet_ChainRecurrence open scoped NNReal
namespace StochApproxDyn.LimitSet
/-- Theorem 5.7 (i) (Benaïm 1999, p. 24): let `Φ` be a semiflow on an arbitrary metric space `M`
and `X : ℝ₊ → M` a precompact asymptotic pseudotrajectory of `Φ`. Then its limit set
`L(X) = ⋂_{t ≥ 0} closure (X([t, ∞)))` is internally chain transitive. -/
theorem limitSet_isInternallyChainTransitive {M : Type*} [MetricSpace M] (Φ : Flow ℝ≥0 M)
(X : ℝ≥0 → M) (hX : IsAsymptoticPseudotrajectory Φ X)
(hpre : IsCompact (closure (Set.range X))) :
IsInternallyChainTransitive Φ (limitSet X) := by sorry
end StochApproxDyn.LimitSet
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.