Natural Orbit-Symmetric Hidden-Vertex Ansatze Fail On M222¶
Claim/Theorem¶
Let M_{2,2,2} be the smallest connected member of the multi-parallel-circuit family, and write
with visible coordinates ordered as
Consider quadratic pseudo-Boolean energies with all pairwise coefficients nonpositive and all coefficients in [-64,64], restricted to either of the following two symmetry-respecting ansatz classes.
Ansatz A: one hidden bit per parallel class¶
Use hidden bits (h_1,h_2,h_3). Require invariance under:
- permutation of the three classes
P_1,P_2,P_3, and - swapping
a_iandb_iinside each class.
Then every such energy is determined by 11 orbit coefficients:
There is no energy of this form whose hidden minimization equals f_{2,2,2}.
Ansatz B: one global hidden bit plus one hidden bit per class¶
Use hidden bits (g,h_1,h_2,h_3), where g is distinguished but the triple (h_1,h_2,h_3) transforms with the three classes. Require the same visible symmetry as above, together with permutation symmetry of the three class-labeled hidden bits.
Then every such energy is determined by 15 orbit coefficients:
There is no energy of this form whose hidden minimization equals f_{2,2,2}.
Equivalently: two of the most natural global hidden-vertex constructions that keep the full pair-permutation symmetry of M_{2,2,2} already fail by exact MILP infeasibility.
This is a derived exact bounded-ansatz obstruction.
- By [[multiple-parallel-classes-on-one-circuit-give-connected-cut-rank-at-least-three-family.md]], the target function depends only on the singleton bit
uand the three pair states, and is invariant under permuting the three pairs and swapping the two elements inside each pair. - For Ansatz A, the symmetry group collapses every quadratic monomial on
(u,a_1,b_1,a_2,b_2,a_3,b_3,h_1,h_2,h_3)to one of the11orbit types listed above. - For Ansatz B, adding a distinguished global bit
genlarges the orbit list to the15types listed above. - In each ansatz, exact realizability reduces to a mixed-integer linear feasibility problem:
- one binary selector chooses the minimizing hidden state for each visible assignment
x; - for every visible-hidden pair
(x,h), one imposes
and the usual selected-state upper bound
- Because all coefficients are restricted to
[-64,64], the linearization is exact once a sufficiently largeMis chosen. The same finite-box logic as in [[smallest-connected-multi-parallel-circuit-member-has-no-1-or-2-auxiliary-realization-in-large-coefficient-box.md]] applies. - For Ansatz A:
- the model has
11continuous coefficient variables and128\cdot 8=1024selector binaries; - HiGHS reports the exact MILP infeasible.
- For Ansatz B:
- the model has
15continuous coefficient variables and128\cdot 16=2048selector binaries; - HiGHS again reports the exact MILP infeasible.
- Therefore neither of these natural orbit-symmetric global hidden-vertex ansatz classes realizes
f_{2,2,2}.
Consequences for the current frontier:
- this closes the first symmetry-respecting global constructive routes beyond the already-failed selector quotient and one-branch-bit local factorization;
- any positive theorem for
M_{2,2,2}must either break these simple orbit-symmetric hidden architectures or use a richer hidden interaction pattern than one per class, even after adding one distinguished global bit; - this still does not prove nonexpressibility of
M_{2,2,2}, because a successful realization could use a less symmetric or more complicated hidden-variable design.
Dependencies¶
- [[multiple-parallel-classes-on-one-circuit-give-connected-cut-rank-at-least-three-family.md]]
- [[smallest-connected-multi-parallel-circuit-member-has-no-1-or-2-auxiliary-realization-in-large-coefficient-box.md]]
- [[branch-min-route-for-connected-family-fails-at-local-submodularity.md]]
Conflicts/Gaps¶
- This node does not prove that
f_{2,2,2}is not hidden-vertex graph-cut representable. - It proves only that two natural symmetry-respecting bounded ansatz classes fail exactly.
- A less symmetric construction, or a more elaborate hidden architecture, remains open.
- Even a future explicit realization or nonexpressibility theorem for this connected family would still need a separate bridge to routing-style
CD(T_n,G)semantics.
Sources¶
10.1016/j.dam.2009.07.00110.48550/arXiv.2109.14599