Remark. The idea of the proof is that the minimum bisection problem is a constrained version of the ▇▇▇▇▇▇▇▇▇▇▇-▇▇▇▇▇▇▇▇▇▇▇ model, which is a spin model where all the spins are independent (cf. ▇▇▇▇▇▇▇▇▇▇▇ and ▇▇▇▇▇▇▇▇▇▇▇, 1975). In the minimum bisection problem, it is required that the partition of the graph be balanced, or equivalently rephrased in spin model terms, it is required that there is the same number of up-spins as down-spins. Therefore, the only difference between the two problems is the solution space. More precisely, we have CMBP ⊂ CSK. Hence ZMBP(β) = Σ e−βR(c,X) ≤ Σ e−βR(c,X) = ZSK(β), (A.2) c∈CMBP
Appears in 2 contracts
Sources: Posterior Agreement for Large Parameter Rich Optimization Problems, Posterior Agreement for Large Parameter Rich Optimization Problems