Genetic Algorithms by Hand: Selection, Crossover, Mutation & Fitness

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Decode 5-bit chromosomes into numbers x, score them with f(x) = x(40 − x), decode gene cells with place values and read the weight of two backpack chromosomes.
• Compute weight, value and excess weight of six backpack chromosomes, turn them into a penalised fitness and find the best pack.
• Turn six fitness values into slices of a roulette wheel, spin it six times with given random numbers, look up the winners with MATCH and draw the wheel as a pie chart.
• Run six 2-way tournaments with given random numbers: draw two entrants, compare their fitness, keep the winner and measure how much fitter the mating pool is than the population.
• Pick cut points from given random numbers, cross three pairs of 5-bit parents with LEFT and RIGHT, then cross two backpack parents gene by gene with IF and weigh the child.
• Mutate four 5-bit children with a given table of random numbers, decode the results, count the flips and compare them with the expected number and the chance of no change.
• Run one complete GA generation for four 5-bit chromosomes: roulette selection with MATCH, crossover with a partner slot, mutation with BITXOR masks, and the best and average fitness of the new generation.
• Chart how best and average fitness converge over 10 generations, enumerate all 32 values of x to find the true optimum, and check when the GA first reached it and how much work it needed.