Completions of one mode word with a prescribed histogram
Provedmme_stothers_general_mode_joint_table_fiber_cardCompletions of one mode word with a prescribed joint histogram.
Fix an integral ten-class profile , a scale , a marginal-supported address of length , a mode , and a -cell histogram whose three grade marginals are the prescribed ones, for every mode and grade . Then the number of marginal-supported addresses that agree with on the -th mode word and have joint histogram exactly is
Fixing the -th word freezes, for each grade , which positions carry that grade in mode ; a completion is then a choice, independently over the nine grades, of how to distribute those positions among the supported triples lying over in mode , with the multiplicities prescribed by . That is a product of nine multinomial coefficients, and regrouping the denominators over all cells gives the displayed quotient. The marginal hypothesis on is exactly what makes each of those nine multinomials well posed, and also what forces the resulting address to be marginally regular in the other two modes.
This quotient is the star degree at ; comparing it with the star degree at the maximum-entropy histogram on the same marginal fibre is what produces the combination loss of Equation (3.4).
import Definitions.Def_mme_stothers_general_outer_profile open MME BigOperators set_option autoImplicit false
theorem mme_stothers_general_mode_joint_table_fiber_card
(base : Fin 10 → ℕ) (m : ℕ)
(a : MME.StothersFourth.GenMarginalSupportedAddress base m) (i : Fin 3)
(k : MME.StothersFourth.GenHashJointMultiplicityTable)
(hkMarginal : ∀ l : Fin 3, ∀ j : Fin 9,
(∑ sigma : {sigma : MME.StothersFourth.GenHashSupportTriple //
sigma.1 l = j}, k sigma.1) =
MME.StothersFourth.genMarginalCount base m j) :
Nat.card
{b : MME.StothersFourth.GenMarginalSupportedAddress base m //
b.1 i = a.1 i ∧
MME.StothersFourth.genHashJointTable b = k} =
(∏ j : Fin 9,
(MME.StothersFourth.genMarginalCount base m j).factorial) /
∏ sigma : MME.StothersFourth.GenHashSupportTriple,
(k sigma).factorial := by
sorry