1 IN2012 Neural and Evolutionary Computing Coursework 2 For this coursework, I would like you to carry out an exploratory investigation on some aspect of a genetic algorithm. This investigation only needs to concentrate on one part of a GA, here are some suggestions: 1. Representations – is binary necessarily the best way to represent the values of the problem you are trying to find a solution to? 1. Number representations. What about trinary? Quatenary? Denary? What about binary numbers themselves? 2. Is binary the best way for a GA? Consider counting in binary: say we have to optimise a binary number, the binary value 0111 is one-less than 1000 which is a problem for a GA because the first string has nothing in common with the second. 3. What about chromosomes? GAs use what are called haploid chromosomes, but could they do better with diploid chromosomes? How would it work? 2. Selection: roulette wheel is regarded as performing less well than tournament selection, but is it true? Also what sort of tournament size work best? Is it related to population size? 3. Crossover: is one-point crossover best? Should crossover be carried out differently, e.g. only at “gene” boundaries. Should representations include “noise” to improve performance – or does it make it worse? 4. Mutation: what is the optimum mutation rate? Should it vary during the run (e.g. sunspot model, Simulated Annealing model – gradually reducing the mutation rate). 5. Populations: What is an optimal population size – if there is such a thing? Should populations be sub-divided? How do you carry out the analysis? An example of a Study I have chosen to investigate the effect of mutation on a genetic algorithm. I now have to mistake a decision: do I compare different (fixed) rates of mutation, or do I compare fixed against variable rates of mutation? 1. A Comparison of fixed rates of mutation. Research Question: Is there one single optimum mutation rate that works for all problems or does it depend on the…