Acquired Intelligence and Adaptive Behaviour 2024 Coursework No 2: Technical report
Acquired Intelligence and Adaptive Behaviour 2024 Coursework No 2
Deadline: 21th of May, 16:00 hrs
Extension: 6-14 pages + references (and appendix if needed)
Submission format: PDF
You have to write a technical report describing an investigation related to the topics covered in class. For choosing the specific topic of investigation, you have to make a number of choices described below.
1. Environment:
You can choose to work with one of three possible setups: a) Physical robotics.
b) Robot simulation, considering details of the physical environment. c) An abstract, simplified environment.
Please note that (a), (b) and (c) are ordered according to their difficulty (with (a) being the most difficult and (c) the easiest). Marking will consider the challenges involved with each scenario, being more generous for more difficult setups and stricter with simpler setups.
2. Agent(s):
You have to choose what type of agent/agents you will study:
-
You can consider a scenario with a single agent, or one with multiple interacting agents.
-
You need to define the “anatomy” of the agent, including what sensors it has, what type
of actions it can do, and what is the internal structure connecting these (i.e. its brain).
-
If you do evolutionary processes, you need to define the genome of the agent and its
encoding.
3. Adaptation mechanisms (algorithm):
You have to choose one or more mechanisms that will let the agent to adapt to its environment:
-
Evolution — GAs or other stochastic methods (particle filters, etc).
-
Learning — back propagation, reinforcement learning, Hebbian learning, etc.
-
A combination of methods involving both evolution and learning.
-
Other methods (speak with the TAs).
Research topic:
After these choices have been made, you have to decide what is the specific topic of your investigation. You can investigate a question that you find most interesting, related to what we have seen in class. If you need inspiration, below you can find some ideas:
-
Compare the performance of various types of GA’s for a given task.
-
Train a deep neural network using different types of GA and compare against back-
propagation.
-
Study the effect of different genotype-phenotype encodings.
-
Evaluating the impact of different choices of hyperparameters in various algorithms.
-
Evaluating sensor/input configuration.
-
Compare performance of algorithms of evolution vs learning for the same task.
-
Design an evolutionary system where learning emerges.
-
Design a scenario where learning feeds back into evolution.
-
Evaluate how behaviour becomes more complex by putting multiple evolving/learning
agents to interact with each other.
-
Other ideas...
Please feel free to search for your own topic. When choosing a topic, please make sure to do the following:
-
Identify a specific hypothesis that you want to test.
-
Perform some analysis to test the hypothesis.
-
Use the results of your analysis to reach a conclusion.
Report structure:
Here is a suggested structure for the report:
-
Introduction: a general explanation of the investigation.
-
Hypothesis and methods: a description and explanation of your choices for
environment, agent, and adaptation mechanism, together with a clear presentation of
your driving hypothesis.
-
Results: a description of the experiments you did and what you found.
-
Discussion: a presentation of your conclusions related to your hypothesis. You can
mention implications of these conclusions, and also describe possible next steps for future work that could eventually continue what you did.
5.
Notes:
-
- There is an infinite number of potential investigations. Different students are expected to
develop non-overlapping topics (please don’t copy what others are doing, it is very easy
to note).
-
- You can use different AI tools (e.g. Chat-GPT) to help yourself with the writing style of
parts of your report. If you use them, you must acknowledge at the end to what degree different sections used such tools. Failure to acknowledge this (which is also easy to note) will be severely penalised.
-
- Write as concisely as you can. If you can fit your ideas into 6 pages that is great; marks are not related to length per se.
Marking scheme
As a guide, the contributions of different aspects of the report run as follows:
-
Technical (30): The quality of your code and the algorithms you present.
-
Presentation (20): The quality of writing and organization of the submitted
document, the quality of the figures and diagrams.
-
Context (20): The extent to which you have motivated the work and discussed the
results in the context of the ideas presented in the lectures.
-
Research (30): The extent to which you've gone beyond the lecture material and
brought in ideas from the course reading, from other sources and your own ideas. The marking criterion will closely follow what you should expect for the 3rd year project.
-
90% – 100%: A truly outstanding project. The project outcomes (system, theory, empirical evaluation) should be essentially faultless, well-structured and carefully tested, proved or rigorously evaluated. There should be full achievement of objectives and evidence of original thought. The project objectives must be very demanding and there should be a wide range of cogently-justified project extensions. The report should be superbly organised and presented and lucidly written. The quality of the research and report should be equally high. The work should be of publishable quality in a peer-reviewed national conference.
-
80% – 89%: An outstanding project. The project outcomes (system, theory, empirical evaluation) should be essentially faultless, well-structured and carefully tested, proved or rigorously evaluated. There should be full achievement of demanding objectives and evidence of original thought. The report should be well organised and presented and clearly written.
-
70% – 79%: Students will show an understanding of all aspects of the project material, producing work without significant error or omission. Project objectives should be reasonably demanding and fully achieved. The report should display excellent organisational and presentational skills, and contain a thorough evaluation and objective critical reflection.
-
60% – 69%: The project should be competent in all respects. The project's primary objectives are somewhat demanding and should be substantially achieved to a reasonable standard. Students will show an understanding of the technical and professional issues involved. The presentation and organisation of the report should be clear.
-
50% – 59%: The project should be competent in most respects. The project objectives may not be very demanding but should be achieved to a reasonable standard. The presentation and organisation of the report should be reasonably clear. There may be some signs of weakness, but overall the grasp of the topic should be sound.
-
40% – 49%: The project will indicate a basic understanding of the methods to be used and how to organise and present the work in the report, but will not have gone beyond this, and there may well be signs of confusion about more complex
material. There should be fair work towards the project objectives and the final
report must clearly represent a development of the interim report.
-
30% – 39%: There should be work towards the project objectives, but significant
issues are likely to be neglected. There may be significant errors or misconceptions in the project. The final report may represent little progress with respect to the interim report.
-
15% – 29%: The project may contain some correct and relevant material, but most issues are neglected or are covered incorrectly. There should be some signs of appreciation of the project requirements.
-
0% – 14%: Very little or nothing that is correct and relevant and there is no real appreciation of the project requirements.
