TorchLean API

FloatLib.Numerics.Enclosure.Rational.Convergence

Convergence through rational interval operations #

Repeated squaring preserves endpoint convergence to a nonnegative target. Together with the containment lemmas, this lets an analytic kernel reduce its argument without losing either soundness or arbitrary accuracy.

theorem FloatLib.Numerics.RationalInterval.tendsto_squareRepeat_lo {α : Type u_1} {l : Filter α} {intervals : αRationalInterval} {x : } (hlo : Filter.Tendsto (fun (a : α) => (intervals a).lo) l (nhds x)) (hx : 0 x) (n : ) :
Filter.Tendsto (fun (a : α) => ((intervals a).squareRepeat n).lo) l (nhds (x ^ 2 ^ n))

The lower endpoints of a repeatedly squared enclosure approach the corresponding power.

theorem FloatLib.Numerics.RationalInterval.tendsto_squareRepeat_hi {α : Type u_1} {l : Filter α} {intervals : αRationalInterval} {x : } (hhi : Filter.Tendsto (fun (a : α) => (intervals a).hi) l (nhds x)) (n : ) :
Filter.Tendsto (fun (a : α) => ((intervals a).squareRepeat n).hi) l (nhds (x ^ 2 ^ n))

The upper endpoints of a repeatedly squared enclosure approach the corresponding power.