Improvement heuristics II - Simulated Annealing

14 important questions on Improvement heuristics II - Simulated Annealing

What is the idea behind SA?

  • Generate a neighbour B from solution A;
  • When the change is positive, accept B;
  • If else, accept B with a certain probability;

How do you calculate the probability of accepting a neighbor in SA?

Boltzmann distribution

Consider the Boltzmann distribution
  • What does delta stand for?
  • What does T stand for?
  • Consider delta when is the probability of acceptance higher?

  • Delta is the difference between the neigbor solution and the current solution;
  • T is the temperature, cooling parameter;
  • The smaller the delta the higher the probability of accepting;
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

How is the temperature updated?

Through a cooling scheme

Explain what the markov chain length does in SA

The temperature changes after M iterations in which M is the markov chain length.

What is a static vs a dynamic Markov chain?

Static
Fixed M (Markov Chain Length)

Dynamic
Increase M after each chain because the probability of transition decreases

SA makes use of a stopping criteria. What are [3] possible termination criteria?

  • Reaching maximum number of iterations;
  • Solution at the end of the Markov Chain did not change for x chains;
  • Temperature reaches a preset minimum value;

What do finite time implements of SA lead to?

Local optima.

Implementation of SA requires [3] things in practice, what are these [3] things?

  1. Well-chosen problem representation;
  2. Transition mechanism;
  3. Cooling Schedule;

In SA what is meant with the transition mechanism, and what are [2] requirements in picking one?

Picking of a suitable neighbor operator.
  • Every solution must be reachable;
  • Must be possible in as few transitions as possible; 

Give [4] transition mechanism examples;

  • K-opt;
  • Or-opt;
  • Swap;
  • Move;

How can the SA algorithm be sped up?

By not evaluating each neighbour solution from scratch but for example, using operator knowledge.

What is the acceptance ratio used for?

The acceptance ratio can be used to choose the initial temperature such that every transition is possible, ensuring high diversification.

In choosing the parameters for the cooling scheme how do you determine Markov Chain Length M and decrease factor alpha and in what order?

  1. Length Markov Chain is the number of neighbor solutions;
  2. Decrease factor alpha follows from the Markov Chain Length;

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo