Let us now turn our attention to the symmetries of the full scenario. We first recall the definitions of part 1.
[1]:
addpath([pwd, '/../../../external/replab']);
replab_init('verbose', 0);
outputGroup = replab.S(2);
outputRep = outputGroup.naturalRep;
inputGroup = replab.S(2);
ioGroup = inputGroup.wreathProduct(outputGroup);
ioRep = ioGroup.imprimitiveRep(outputRep);
replab_init: Initialization done.
The same story repeats for relabelings of parties: the scenario involves two homogeneous parties. We thus have two copies of the group relabeling inputs and/or outputs (one for Alice, one for Bob), and a copy of \(S_2\) that permutes the parties. This group desribing all possible relabellings in this scenario is thus given by:
[2]:
scenarioGroup = outputGroup.wreathProduct(ioGroup);
[3]:
scenarioGroup.sample
ans =
{
[1,1] =
1 2
[1,2] =
{
[1,1] =
{
[1,1] =
2 1
[1,2] =
{
[1,1] =
1 2
[1,2] =
1 2
}
}
[1,2] =
{
[1,1] =
2 1
[1,2] =
{
[1,1] =
1 2
[1,2] =
2 1
}
}
}
}
The representation on the behavior \(P(ab|xy)\) is however a primitive representation, as \(P(a|x; b|y)\) ressembles a tensor. Inside each party, we use the imprimitive representation constructed before.
[4]:
probRep = scenarioGroup.primitiveRep(ioRep);
Decomposition of the full probability space
We can now compute the decomposition of this representation on \(P(ab|xy)\):
[5]:
dec = probRep.decomposition
dec =
Orthogonal reducible representation
dimension: 16
field: 'R'
group: Finite group of order 128
isUnitary: true
parent: Orthogonal representation
component(1): Isotypic component R(1) (trivial)
component(2): Isotypic component R(1) (nontrivial)
component(3): Isotypic component R(2) (nontrivial)
component(4): Isotypic component R(4) (nontrivial)
component(5): Isotypic component R(4) (nontrivial)
component(6): Isotypic component R(4) (nontrivial)
There are 6 irreducible components
[6]:
dec.nComponents
ans = 6
These components correspond to the following physical elements ( see arXiv:1610.01833 ) for more details:
The global normalization of probabilities
The difference of normalization between \(x=y\) and \(x\ne y\)
The difference of normalization between \(x=0\) and \(x=1\)
The signaling between Alice and Bob
The marginal probabilities of Alice and Bob
The correlation between Alice and Bob