07 - Stochastic Optimisation

How long did you work on the last exercise sheet?

• I didn't do it.

• < 5 hours.

• 5-8 hours.

• > 8 hours.

What is the objective of simulated annealing?

• Find a good exploitation-exploration tradeoff automatically.

• Find all local optima.

• Reduce the temperature of a system.

• Find the global optimum.

• 1

• 0

In terms of a minimization problem, low energy states should have a [...] probability of occurrence.

• [ low ]

• [ high ]

Convergence to the global optimum is guaranteed if the temperature decreases no faster than log(t). But ...

• we get round off errors.

• It depends. Are we maximizing or minimizing?

• it's too slow.

• left

• right

• middle

• none

• more uniform.

• peaky.

• gaussian.

• flatter.

Minimization: The global minimum should be a state with [...] energy.

• [lowest energy]

• [highest energy]

• with high probability.

• with low probability.

• always.

• with 50/50 % chance.

In mean-field annealing, we are interested in the [...] moments.

• [first]

• [second]

• [all higher-order]