COMP5400: Biological and Bio-Inspired Computation
Coursework 1:
Evolution and Co-evolution with BEAST
Due Date: 1 March, 2023
Module Leader: Netta Cohen
Marking
This coursework is summative and is worth 40% of the total grade for this module.
Assessment
Please submit this coursework as a single tar-gzip archive with extension “.tgz” or extension “.tar.gz” via
Turnitin. You will need to submit:
A declaration that the work is your own. You will be asked to share your long simulations for Q6.
Make sure any data you share has your name on it and that you properly credit any data you obtain
from your colleagues.
A pdf of typed answers to questions. Hand-written answers will not be accepted. Please use arial
fonts with size 11pt. Figures must be of sufficient size and clearly labeled (including e.g., axis label and
ticks, in a legible font, min. size 9pt).
Scripts that you used to generate plots for Questions 1 and 6. Without these scripts, you may be
awarded no marks for these questions.
The modified mouse.cc code for Q4. If the code does not compile and run properly, you may be
awarded no marks for this question.
The specific code modifications (not the entire file) that you made to answer Q4 should also be
included in your answer sheet to Q4. Without being able to inspect these modifications, you may be
awarded no marks for Q4. To include these modifications: please cut and paste the relevant lines of
code into your answer sheet, rather than using screen shots. Screenshots with a black background or
with any text colour that is not clearly legible in the pdf of your answer sheet will be penalised.
Keep your answers to question clear and well-reasoned. You will not get additional points for longer
answers where those are not adding value to your response.
Plagiarism
You must compile and run your own code. Code modifications, scripts and answers to questions must be
prepared by yourself. It is strictly prohibited to borrow or share code. You may help each other with BEAST
installation. You may discuss what the demos are about. You may also discuss material covered lectures, your
related reading and ideas relevant to the coursework together with your peers, but you may not discuss or
share answers to questions.
Questions
Part 0 – Formative: This part of the assignment is optional, but encouraged. You are encouraged to
work with your classmates. The purpose of this component is to acquaint you with BEAST, and to
introduce you to Braitenberg vehicles (which we will study in detail in later lectures).
1. Run the Braitenberg demo in Beast. For these questions you need not look at the code, or look
any material up. Just watch the demo: you will see agents roaming around the arena. What are they
doing? Do they have purpose? If so, try to articulate it. Do you think these agents have an artificial
brain? What evidence do you have for your conclusion (either way).
The remainder of the worksheet is summative.
Part I: Run the mouse demo in BEAST.
1. Plot the fitness of the mice as a function of the generation count. Be judicious where you
terminate your simulation. Don’t stop too early. Briefly motivate why you show the particular range
you chose.
[5 marks]
2. Describe the behaviour of the mice as it evolves over evolutionary time.
[5 marks]
3. Describe the fitness function used by the mice. It can be found in the file ‘mouse.cc’. Experiment
with the fitness function: change it and run the simulation again. Judging from the behaviour of the
mice, explain whether you believe that the performance of the mice improves or deteriorates with
your changes. Look in the code to see what parameters the genetic algorithm uses. Consider the
influence of these parameters in your answer.
[10 marks]
4. How would you define the performance objectively and how would you evaluate the relative
performance of the different fitness functions? Implement and evaluate them. Also experiment
with different sensor configurations. Explain the reasoning behind your experiments.
[20 marks]
5. Based on the lecture notes and the literature cited there, discuss what you understand by the
term collective behaviour. Describe and explain the behaviour of the mice. Do you see evidence
of collective behaviour? Explain your answer.
[15 marks]
Part II: Run the chase demo in BEAST. Run three demonstrations for 100 generations, each. (You can
do this in parallel.) Record the behaviour of both types of agent after 10, 50 and 100 generations. Let
one simulation run for much longer (several thousands of generations).
6. Plot the fitness of each agent as a function of the generation count for both types of agent
for all runs. Think of a good way to represent and compare your results in plots. Include at least two
of your colleagues’ results from the long simulation. To give proper credit to others, you must state
the full names of students who have shared data with you.
[10 marks]
7. Describe the behaviour of the agents. Is this an example of co-evolution? Argue. You should
explain your reasons and evidence your answer (for co-evolution or for its absence) based on the
behaviour of the agents after different numbers of generations.
[10 marks]
8. Judging the fitness plots for the same simulations, do you see evidence for co-evolution? Discuss
and evidence your answers.
[10 marks]
9. Do you consider the behaviour of the agents to be intelligent? Answer extensively, and give your
own motivation and evidence (for intelligence or for its absence, or for inconclusive evidence). In
your answer, consider one or more definitions of intelligence. Some definitions may come from the
scientific literature, for example the paper ‘Intelligence without Representation’ by Rodney Brooks,
which can be found on Minerva. If you wish (note, this is not required) you may propose your own
definition of intelligence in your answer.
[15 marks]