You are welcome to implement both "Binary-tournament selection with replacement" and "Binary-tournament selection without replacement" and compare their takeover time. This is an option, not a requirement.
You can assign epsilon with any value you like for your linear-ranking selection method. However, you need to document it and discuss its impact on the evolutionary behavior.
You will produce two figures for each of the "population average fitness" and "population best fitness" plot.
The first figure resports reslusts from time step 0 to 1,000, each data point is the average of 10 or 20 time steps.
The second figure reports results from time step 800 to 1,000, each data point is the average of 5 or 10 time steps.
The second figure has a different y-scaling from that in the first figure, hence allows you to see performance differences among different PSO paremeter setups.
When discuss about why one PSO parameter setup gives better results than that produced by the other setup, you can analyze exploration vs. exploitation.
For example, does one setup gives more diversity (population dispersion) than the other? Does one setup produce better population average fitness too quickly?
Project proposal (1-2 pages) is due on November 15. The total marks of the project can be either 100/100 or 120/100. If you are interested in the later, you will describe the additional work you will do in your proposal for the bonus marks.