InformationTheory.Shannon.BroadcastChannel.Superposition.TimeShare
Broadcast channel — absorbing a time-sharing variable into the superposition cloud #
A point of the UV outer region is described by a five-tuple law, whose two information slots
I(U; Y₂) and I(X; Y₁ ∣ U) are the corners of a rectangle the superposition inner bound only
reaches after the rate pair has been traded along a segment. Time sharing performs that trade:
the auxiliary is kept with probability lam and collapsed to a constant with probability
1 - lam, and both branches are carried by the single auxiliary Bool × U, whose first
component records which branch was taken.
Because the tag is recoverable from the auxiliary, the mixture identities for the two slots are
exact where they need to be. The receiver-2 corner only needs the branch that keeps the
auxiliary, so the tag's own contribution is discarded and the slot is bounded below by
lam * I(U; Y₂). The satellite slot is an equality, lam * I(X; Y₁ ∣ U) + (1 - lam) * I(X; Y₁),
because collapsing the auxiliary turns the conditional information into the unconditional one.
The second endpoint therefore contributes I(X; Y₁), and the segment traced by lam stays above
the outer point as soon as the rate pair also satisfies max R₁ 0 + R₂ ≤ I(X; Y₁). That sum
constraint is all the assembly asks for beyond the three outer inequalities, and it comes for free
over a less noisy channel, which forces I(X; Y₁ ∣ U) + I(U; Y₂) ≤ I(X; Y₁) at every law.
Main definitions #
uvCloudLaw νanduvSatelliteKernel ν— the achievability pair read off a five-tuple law.uvTagTrue ν,uvTagFalse σ,uvMixKernel ν σanduvMixLaw ν σ lam— two channel laws mixed along a Bernoulli tag carried by the auxiliary, each of them keeping its auxiliary under its own tag.uvCollapse ν u₀anduvTagConst ν u₀— the law with its auxiliary collapsed to the constantu₀, and its false-tagged copy.uvTimeShareLaw ν u₀ lam— the law and its own collapsed copy mixed with weightlam.boolProdAuxEquiv m— the tagged auxiliary alphabet re-encoded intoMarton.bcAuxAlphabet.
Main statements #
bcInfo₂_uvCloudLaw,bcInfo₁_uvCloudLawandbcInfoJoint_uvCloudLaw— the three informations of the achievability pair read off a channel law are the corresponding slots of that law.mul_uvInfo₂_le_uvInfo₂_uvMixLawandmul_condMutualInfo_le_condMutualInfo_uvMixLaw— the two slots of a mixture of two channel laws against the slots of the branch it keeps.mul_uvInfo₂_le_uvInfo₂_uvTimeShareLawandcondMutualInfo_uvTimeShareLaw— the two slots of the time-shared law against the slots of the original one.uvInfoJoint_uvTimeShareLaw— time sharing leaves the input-output informationI(X; Y₁)unchanged.exists_bcInfo_ge_sumRate_of_isUVChannelLaw— a rate pair satisfying the three UV outer inequalities of a channel law together with the sum constraintmax R₁ 0 + R₂ ≤ I(X; Y₁)is dominated by the three informations of some achievability pair, with no comparison between the two receivers.exists_bcInfo_ge_of_lessNoisy_of_isUVChannelLaw— over a less noisy channel, a rate pair satisfying the three UV outer inequalities of a channel law is dominated by the two informations of some achievability pair, the sum constraint coming for free from the comparison.
Implementation notes #
uvMixLaw mixes two arbitrary channel laws along the tag rather than a law and its own collapsed
copy, because the same construction carries the perturbation toward the uniform law as well:
forgetting the tag turns uvMixLaw ν (uvUniformLaw W v₀) lam into uvPerturbLaw W ν v₀ lam.
Time sharing is the case where the second law is uvCollapse ν u₀, so uvTimeShareLaw is defined
as that specialization and both mixtures share the lemmas about the tag.
Mixing two channel laws along a Bernoulli tag #
The mixture of two tagged laws #
InformationTheory.Shannon.BroadcastChannel.uvTagTrue
sourceThe branch that keeps the auxiliary, tagging it with true.
Equations
Instances For
Used by
InformationTheory.Shannon.BroadcastChannel.uvTagFalse
sourceThe branch that keeps the auxiliary, tagging it with false.
Equations
Instances For
Used by
InformationTheory.Shannon.BroadcastChannel.uvTagTrue_isProbabilityMeasure
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvTagFalse_isProbabilityMeasure
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvMixKernel
sourceThe kernel that selects one of two tagged laws from the tag.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Used by
InformationTheory.Shannon.BroadcastChannel.uvMixKernel_isMarkovKernel
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvMixLaw
sourceThe two laws mixed with weight lam, carried by the auxiliary Bool × U.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Used by
InformationTheory.Shannon.BroadcastChannel.uvMixLaw_isProbabilityMeasure
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvMixLaw_eq
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvMixKernel_ae_tag
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvTagTrue_map_forget
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvTagFalse_map_forget
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvMixLaw_map_forget
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvMixLaw_isUVChannelLaw
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvMixLaw_map_tag
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvInfo₂_uvTagTrue
sourceUsed by
Time sharing as the mixture with the collapsed law #
InformationTheory.Shannon.BroadcastChannel.uvCollapse
sourceThe law with its auxiliary collapsed to the constant letter u₀.
Equations
Instances For
Used by
InformationTheory.Shannon.BroadcastChannel.uvCollapse_isProbabilityMeasure
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvTagConst
sourceThe branch that collapses the auxiliary to the constant (false, u₀).
Equations
Instances For
Used by
InformationTheory.Shannon.BroadcastChannel.uvTagConst_isProbabilityMeasure
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvTagConst_eq_map
sourceUsed by
The information slots of a mixture and of the time-shared law #
The receiver-2 corner #
InformationTheory.Shannon.BroadcastChannel.mul_uvInfo₂_le_uvInfo₂_uvMixLaw
sourceUsed by
The satellite conditional information #
InformationTheory.Shannon.BroadcastChannel.condMutualInfo_uvTagTrue
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.mul_condMutualInfo_le_condMutualInfo_uvMixLaw
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.condMutualInfo_uvTagConst
sourceUsed by
Reading an achievability pair off a channel law #
The cloud law and the satellite kernel #
InformationTheory.Shannon.BroadcastChannel.uvCloudLaw
sourceThe cloud law of a five-tuple law: the marginal of the first auxiliary.
Equations
- InformationTheory.Shannon.BroadcastChannel.uvCloudLaw ν = MeasureTheory.Measure.map (fun (q : U × V × α × β₁ × β₂) => q.1) ν
Instances For
Used by
InformationTheory.Shannon.BroadcastChannel.uvSatelliteKernel
sourceThe satellite kernel of a five-tuple law: the conditional law of the input letter given the first auxiliary.
Equations
Instances For
Used by
InformationTheory.Shannon.BroadcastChannel.uvCloudLaw_isProbabilityMeasure
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.uvSatelliteKernel_isMarkovKernel
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.bcJointDistribution_uvCloudLaw
sourceUsed by
The three informations of the achievability pair #
InformationTheory.Shannon.BroadcastChannel.bcInfo₂_uvCloudLaw
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.bcInfo₁_uvCloudLaw
sourceUsed by
InformationTheory.Shannon.BroadcastChannel.bcInfoJoint_uvCloudLaw
sourceUsed by
Landing in the auxiliary alphabet of the inner bound #
InformationTheory.Shannon.BroadcastChannel.boolProdAuxEquiv
sourceThe tagged auxiliary alphabet re-encoded into the auxiliary alphabet of the inner bound.