The probability of making the transition from the current state <math>s</math> to a candidate new state <math>s'</math> is specified by an acceptance probability function <math>P(e, e', T)</math>, that depends on the energies <math>e = E(s)</math> and <math>e' = E(s')</math> of the two states, and on a global time-varying parameter <math>T</math> called the temperature. States with a smaller energy are better than those with a greater energy. The probability function <math>P</math> must be positive even when <math>e'</math> is greater than <math>e</math>. This feature prevents the method from becoming stuck at a local minimum that is worse than the global one. | The probability of making the transition from the current state <math>s</math> to a candidate new state <math>s'</math> is specified by an acceptance probability function <math>P(e, e', T)</math>, that depends on the energies <math>e = E(s)</math> and <math>e' = E(s')</math> of the two states, and on a global time-varying parameter <math>T</math> called the temperature. States with a smaller energy are better than those with a greater energy. The probability function <math>P</math> must be positive even when <math>e'</math> is greater than <math>e</math>. This feature prevents the method from becoming stuck at a local minimum that is worse than the global one. |