Diamond distance between quantum channels
import Quantum_state import Quantum_channel import Choi_correspondence import Trace_distance /-- are the channel outputs close for all inputs? -/ def diamond_dist [quantum_channel π© β³] : β := Sup { Sup { β₯1ββπ© Ο - 1βββ³ Οβ₯β | quantum_state Ο } | n : β } notation `β₯π© - β³β₯β` := diamond_dist π© β³ Ο is state over main system A and reference system of dim n. /-- "The normalized diamond distance can be computed efficiently by means of the following semi-definite program." -/ theorem diamond_dist_sdp [quantum_channel π© β³] : 1/2 * β₯π© - β³β₯β = Inf { pTr_B β₯ZRBβ₯β | ZRB β₯ 0 β§ ZRB β₯ Choi π© - Choi β³ } := sorry
βPrepare a state ΟRC of a reference system R and the channel input system C. Feed system C into the unknown channel. Perform a POVM measurement {ΞRD i} on the channel output and R.β
import POVM /-- "The maximum possible error between π©CβD and β³CβD; Optimizing with respect to all preparations and measurements, is equal to the normalized diamond distance:" -/ theorem sup_prob_diff_eq_diamond_dist : Sup { β£Tr((Ξ x) 1βπ©(ΟRC)) - Tr((Ξ x) 1ββ³(ΟRC))β£ | quantum_state ΟRC β§ POVM Ξ } = 1/2 * β₯π© - β³β₯β := sorry
Is diamond_dist
suitable as a metric of a metric space?
theorem diamond_dist_self : β [quantum_channel π©], β₯π© - π©β₯β = 0 := sorry theorem diamond_dist_comm : β [quantum_channel π© β³], β₯π© - β³β₯β = β₯β³ - π©β₯β := sorry theorem diamond_dist_triangle_inequality : β₯π© - β³β₯β β€ β₯π© - Ξβ₯β + β₯Ξ - β³β₯β := begin -- can this be proven? -- rw diamond_dist, end