Branch-Min Route For Connected Family Fails At Local Submodularity¶
Claim/Theorem¶
Let M be a member of the connected multi-parallel-circuit family from [[multiple-parallel-classes-on-one-circuit-give-connected-cut-rank-at-least-three-family.md]], with ground set
For a cut L\subseteq E, write
Define
and
Equivalently, U(L) counts the classes P_i that are not monochromatic with the same cut value as the singleton element u.
Then the connectivity function from [[multiple-parallel-classes-on-one-circuit-give-connected-cut-rank-at-least-three-family.md]] satisfies the exact branch formula
So there is a natural candidate hidden-vertex realization route using one global branch bit h and local per-class gadgets
where p=\sum_{e\in P} x_e and m=|P|, because then formally
However, for every m\ge 2, the visible local gadget \psi_m is not submodular. Consequently, by [[minimizing-hidden-binary-submodular-energy-preserves-submodularity.md]], \psi_m is not hidden-vertex graph-cut expressible. Therefore this whole branch-min route cannot be implemented by independent per-class hidden-vertex gadgets controlled only by one global branch bit.
This is a derived obstruction statement.
-
By [[multiple-parallel-classes-on-one-circuit-give-connected-cut-rank-at-least-three-family.md]],
\[ \lambda_M(L) = S(L) + (1-\varepsilon_L)\,\mathbf 1[\exists i:\ p_i(L)=m_i] + \varepsilon_L\,\mathbf 1[\exists i:\ p_i(L)=0]. \] -
If
\varepsilon_L=0, let\[ F(L):=\#\{i:\ p_i(L)=m_i\}. \]Then
U(L)=S(L)+F(L), while\[ \lambda_M(L)=S(L)+\mathbf 1[F(L)\ge 1]. \]Hence
\[ \lambda_M(L)=\min\bigl(S(L)+F(L),\,1+S(L)\bigr)=\min\bigl(U(L),\,1+S(L)\bigr). \] -
If
\varepsilon_L=1, let\[ E(L):=\#\{i:\ p_i(L)=0\}. \]Then
U(L)=S(L)+E(L), while\[ \lambda_M(L)=S(L)+\mathbf 1[E(L)\ge 1]. \]So again
\[ \lambda_M(L)=\min\bigl(S(L)+E(L),\,1+S(L)\bigr)=\min\bigl(U(L),\,1+S(L)\bigr). \]This proves the branch formula.
-
The displayed
h-decomposition is just a rewriting of\min(U,1+S)into two branches:h=0yieldsU, andh=1yields1+S. -
Now fix any
m\ge 2, and let\mathbf 1^mdenote the all-ones class assignment. Consider\[ \alpha:=(u,h,x_P)=(0,1,\mathbf 1^m), \qquad \beta:=(u,h,x_P)=(1,0,\mathbf 1^m). \]Then
\[ \psi_m(\alpha)=0, \qquad \psi_m(\beta)=0, \]because the
h=1branch sees a full class and theh=0branch withu=1also sees a class monochromatic withu.But
\[ \alpha\wedge\beta=(0,0,\mathbf 1^m), \qquad \alpha\vee\beta=(1,1,\mathbf 1^m), \]so
\[ \psi_m(\alpha\wedge\beta)=1, \qquad \psi_m(\alpha\vee\beta)=0. \]Therefore
\[ \psi_m(\alpha)+\psi_m(\beta)=0<1=\psi_m(\alpha\wedge\beta)+\psi_m(\alpha\vee\beta), \]which violates submodularity.
-
By [[minimizing-hidden-binary-submodular-energy-preserves-submodularity.md]], every hidden-vertex binary-submodular realization has a submodular visible function. Since
\psi_mis not submodular, it cannot be hidden-vertex graph-cut expressible. -
Hence the branch formula does not directly give a connected positive theorem: the obvious implementation with one global branch bit and independent local gadgets is impossible already at the visible local-gadget level.
Consequences for the current frontier:
- the selector quotient is not the only natural route to the first connected
\chi\ge 3family, but this alternative branch-min route also fails in its obvious local-gadget form; - the failure is stronger than a bounded-auxiliary search obstruction: it is an exact visible non-submodularity obstruction for the required per-class branch gadget;
- this still does not rule out a more global hidden-vertex realization of the whole family, because a successful construction could couple the classes in a way that does not factor through independent
\psi_{m_i}blocks.
Dependencies¶
- [[multiple-parallel-classes-on-one-circuit-give-connected-cut-rank-at-least-three-family.md]]
- [[minimizing-hidden-binary-submodular-energy-preserves-submodularity.md]]
Conflicts/Gaps¶
- This node does not prove that the connected multi-parallel-circuit family itself is not hidden-vertex graph-cut representable.
- It rules out only the specific
\lambda_M=\min(U,1+S)implementation that factors through one global branch bit and independent per-class visible gadgets\psi_{m_i}. - A more global hidden-vertex realization that keeps richer cross-class structure remains open.
- Even a future whole-family positive or negative theorem would still need a separate argument to connect hidden-vertex realizability to routing-style
CD(T_n,G)semantics.
Sources¶
10.48550/arXiv.2109.1459910.1016/j.dam.2009.07.00110.1007/s10878-017-0136-yKashyap 2008 preprint: A Decomposition Theory for Binary Linear Codes