Optimisation for Artificial Intelligence, a 4-day course
« Previous | Up ↑ | Next »
Simulated annealing is a trajectory method: we manipulate one element of population at a time and construct a series of $x_i$ in order to make it converge toward an optimum.
Simulated annealing is inspired by metallurgy:
after each iteration, we pick a $x_{i+1}$ “near” $x_i$, and evaluate the variation of energy $\Delta E$:
This exercice may look very easy, but you will realise there is a lot left to your consideration:
And of course, what will work for a type of problem may not be that efficient for other types of problems.
« Previous | Up ↑ | Next »