Evolutionary Computation

Due: Thursday, 26 October 2017 at the beginning of class

- Follow the general homework directions.
- Make sure you cite all your references and contacts.

- Read
- Chapter 8 and 9 in textbook.

- Problems
- Implement the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) in Java. You must write the CMA-ES from scratch. Investigate the CMA-ES optimization capabilities on finding the minimizer of the Rosenbrock 2D landscape with a = 1 and b = 100. Experiment with the population size. Run at least 10 experiments. Determine the average (over 20 runs) take over time, number of generations, and clock time to find the optimzier. Make a graphs of the parameter sets to show how varying the parameters effects the average number number of generations. Make sure to turn in any code needed to run the investigations.
- Now chose another fitness landscape. See how the best parameter sets for the Rosenbrock 2D landscape work on the new landscape. Repeat the same set of investigations on the new landscape. What conclusions can you draw from your investigations?
- Give arguments why mutation strength (e.g., p
_{m}or σ) should be increased during a run. Give arguments why it should be decreased.