§ Computational time vs solution quality – trade off
§ Trivial or inferior moves are allowed for better future solutions hopefully § SA escapes entrapment at local optimum using a probability condition that accepts or rejects an inferior move § No worse move is always accepted § Iterations increase then SA becomes more selective § Temprature (T) based on a schedule § T0 is r0F(s0) where r is between 0-1 § Ti is riTi-1 where i = 1,2,…,I § Sa last accepted move § At iteration k the probability of accepting a neighborhood solution sk+1
1, 𝑖𝑓 𝐹 𝑠)*+ 𝑖𝑠 𝑛𝑜𝑡 𝑤𝑜𝑟𝑠𝑒 𝑡ℎ𝑎𝑛 𝐹(𝑠; )
§ 𝑃 𝑎𝑐𝑐𝑒𝑝𝑡 𝑠)*+ 𝑠)*+ ∈ 𝑁 𝑠) =/ = >(?@ )=>(?ABC ) 𝑒 D 𝑜/𝑤 § Inferior move probability is compared with a random number: § Accept if § Rk is less than or equal to Inferior move probability 3 4 § 𝑥 ∈ 1, … , 8 § 𝑦 ∈ {1, … , 10} § 𝑧 ∈ 1, … , 10 § 𝑥 ≠ 7, 𝑦 ≠ 2, 𝑥 − 𝑦 = 3𝑧 § 800 possible solutions § Reduce solution space § Min x is 4 § Max y is 5 § Max z is 2