You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
* Add plan for #290: [Model] IntegralFlowWithMultipliers
* Add IntegralFlowWithMultipliers model
* Polish IntegralFlowWithMultipliers paper and tests
* chore: remove plan file after implementation
* Fix formatting after merge conflict resolution
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Fix paper solve command: add --solver brute-force (no ILP path for this model)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Given a directed graph $G = (V, A)$, distinguished vertices $s, t in V$, arc capacities $c: A -> ZZ^+$, vertex multipliers $h: V backslash {s, t} -> ZZ^+$, and a requirement $R inZZ^+$, determine whether there exists an integral flow function $f: A -> ZZ_(>= 0)$ such that (1) $f(a) <= c(a)$ for every $a in A$, (2) for each nonterminal vertex $v in V backslash {s, t}$, the value $h(v)$ times the total inflow into $v$ equals the total outflow from $v$, and (3) the net flow into $t$ is at least $R$.
5261
+
][
5262
+
Integral Flow With Multipliers is Garey and Johnson's gain/loss network problem ND33 @garey1979. Sahni includes the same integral vertex-multiplier formulation among his computationally related problems, where partition-style reductions show that adding discrete gain factors destroys the ordinary max-flow structure @sahni1974. The key wrinkle is that conservation is no longer symmetric: one unit entering a vertex may force several units to leave, so the feasible integral solutions behave more like multiplicative gadgets than classical flow balances.
5263
+
5264
+
When every multiplier equals $1$, the model collapses to ordinary single-commodity max flow and becomes polynomial-time solvable by the standard network-flow machinery summarized in Garey and Johnson @garey1979. Jewell studies a different continuous flow-with-gains model in which gain factors live on arcs and the flow may be fractional @jewell1962. That continuous relaxation remains polynomially tractable, so it should not be conflated with the NP-complete integral vertex-multiplier decision problem catalogued here. In this implementation the witness stores one bounded integer variable per arc, giving the direct exact-search bound $O((C + 1)^m)$ where $m = |A|$ and $C = max_(a in A) c(a)$.
5265
+
5266
+
*Example.* The canonical fixture encodes the Partition multiset ${2, 3, 4, 5, 6, 4}$ using source $s = v_0$, sink $t = v_7$, six unit-capacity arcs out of $s$, six sink arcs with capacities $(2, 3, 4, 5, 6, 4)$, and multipliers $(2, 3, 4, 5, 6, 4)$ on the intermediate vertices. Setting the source arcs to $v_1$, $v_3$, and $v_5$ to $1$ forces outgoing sink arcs of $2$, $4$, and $6$, respectively. The sink therefore receives net inflow $2 + 4 + 6 = 12$, exactly meeting the requirement $R = 12$.
caption: [Integral Flow With Multipliers: the blue branches send one unit from $s$ into $v_1$, $v_3$, and $v_5$, forcing sink inflow $2 + 4 + 6 = 12$ at $t$.],
5315
+
) <fig:ifwm>
5316
+
]
5317
+
]
5318
+
}
5319
+
5254
5320
#problem-def("AdditionalKey")[
5255
5321
Given a set $A$ of attribute names, a collection $F$ of functional dependencies on $A$,
5256
5322
a subset $R subset.eq A$, and a set $K$ of candidate keys for the relational scheme $chevron.l R, F chevron.r$,
0 commit comments