Individual Report

The individual report should be no more than 6 pages in length (including figures, but excluding any references). Any format/style can be used, minimum font size of 11. No cover page or content page is needed.

Deadline: 15th June Midnight CET (Submission as a single PDF) on Moodle.

Part 1: Black-Box Optimization (50% of grade, Approximately 3 pages)

1.1 Choose two of the following: Bayesian optimization, simulated annealing, genetic algorithms, particle swarm operator or Tabu Search.

Choose two different 3D ‘functions’ which have different forms

  1. Which is has many different local minimal and maxima (e.g. in Matlab @rastriginsfcn; )

  2. One which has only a few minima/maxima (e.g. @ps_example;)

Figure 1 - left Rastinginsfcn and right ps_example.

For each method you choose optimize one parameter for the two different functions and quantaitively and systematically evaluate how for this parameter, the performance (i.e. fitness function) and efficiency (i.e. convergence time) is affected. Discuss these results and how the optimal parameter is/may be different for the two functions.

Example parameters to investigate include:

1.2 Comparison of the Methods. Discuss which of the methods if best for the two different functions and explain why.

Part 2: PCA on data-set (15% of grade, approximately 1/2-1 page)

Figure 2 - different species plotted against sepal width and sepal length.

We have an example data set of iris flowers.  This has 150 measurements from iris flowers, and for each flower we have four features (sepal length, sepal width, petal length, petal width).  The flowers belong to three different species. We want to find a reduced dimensionality representation of possible flowers. The iris data set can be found here: https://archive.ics.uci.edu/ml/datasets/iris or using matlab(‘load fisheriris;’ after which stores the four measurements for the 150 plants (in meas array) and species stores the label of each species of each of the 150 plants.