COMM510 Multi-Objective Optimisation & Decision Making Assignment

1 Assignment

This coursework requires the submission of a concise document including a background and literature reviewon a research topic you have chosen from the list in Section 2. The submission must also include a descriptionof the research programme you intend to undertake to address the research topic, including aspects such asdetailing the different aspects of the broader research question you will examine, the experimental design, thepresentation of the experiments, the analysis of the results, a discussion and any conclusions, and setting outfuture work directions given these results. This assignment marking criteria are at the end of this document.

The submitted document should demonstrate an understanding of the topic area and the research question.It should present an appropriate programme of research to investigate the research question posed.Precisely, the document (maximum of four pages, excluding references) should include:

• An introduction.
• A review of the literature about the topic, describing the background to the chosen research question andwhat work has been done in the existing literature to investigate it.
• A reasoned plan of the empirical work to be undertaken, including (as suitable), algorithms to compare,test problems to use, quality measures to employ, experimental protocols, how the results will be evaluated,etc.
• Your prior expectations of the results, given any insights from the literature and your understanding ofthe task and research question. This might take the form of a research hypothesis to be investigated.
• A presentation of the results obtained in an appropriate form (e.g. tables, plots, etc.).
• An analysis of the results.
• A contextualisation of the results, relating them to the existing literature.
• A conclusion which outlines potential future research directions that lead on from your work.

2 Topics

Below is the list of research topics for the COMM510 coursework. You must select one topic for this assignment.

1. The impact of initialisation on MOEAs. In this project topic you should explore the researchquestion: how much of the computational budget (objective function evaluations) should be spent oninitialisation? Typically the number of random initial solutions (if no existing solutions are usable), isthe same as the search population – however it is possible to generate substantially more initial randomsolutions and take the “N” best of these to form the initial search population. You will want to explore theeffect of varying the amount of budget you assign to initialisation versus search for one or more optimisers.
2. How amenable/exploitable are test problems suites to seeding? In this project topic you shouldexplore the research question: how much easier are test problems from a suite if search is initialised froma set of solutions which include a Pareto optimal decision vector, and why? You should contrast at least two multi-objective test suites in this project.
3. (1 + 1) or (|A| + 1)? In this project topic you should explore the research question: how (and why)does the relative performance of the simple “greedy” multi-objective evolution strategy optimiser vary if a(1+1) formulation is used, or a (|A|+1) formulation. Where in a (1+1) (as in PAES), the child replacesthe parent if it dominates it, or if is non-dominated with the approximation set but is in a preferredregion (e.g. less dense in objective space), whereas in the (|A|+1) approach the parent newly drawn eachgeneration from the approximation set. You should consider a range of test problems.
4. Influence of different scalarisations in MOEA/D. In this project topic, you should explore theresearch question: What is the effect of different scalarisations on the performance of MOEA/D? Thereare no standard guidelines on selecting a particular scalarising function. You will want to explore theeffect on the performance of MOEA/D with minimum four different scalarising functions.
5. Which MOEA is the best? Comparison of different performance indicators. In this projecttopic you will answer the research question: What is the final ranking of different MOEAs if you usedifferent performance indicators? In the literature, several performance indicators exist. You will want toexplore the correlation between the run time and ranking of different algorithms using different indicators.You can select three indicators and two MOEAs.
6. Looking for decision-maker's desired solution: Comparison of different preference basedMOEAs. In this project topic, you should explore the research question: Which preference based algorithmis the best considering decision-maker's preferences? It is desirable to find a (set of) solution(s)preferable to the decision-maker. Different elements in MOEAs e.g. selection criterion have been adaptedin MOEAs for this task. In this project, you will explore the run-time performance of different preferencebased MOEAs with desirable objective function values (also known as aspiration level) as the preferences.You can use three MOEAs for comparison.
7. Looking for decision-maker's desired solution: Comparison of different scalarising functionsbased algorithms: In this project topic, you should explore the research question: Which scalarisingfunction based algorithm is the best considering decision-maker's preferences? It is desirable to find a(set of) solution(s) preferable to the decision-maker. Many methods use scalarising functions to considerdecision-maker's preferences. In this project, you will explore the run-time performance of different scalarisingfunctions with desirable objective function values (also known as aspiration level) as the preferences.You can use three scalarising functions for comparison.
8. Mono or Multi? Which is better in Multi-objective Bayesian optimisation: In this projecttopic, you should explore the research question: how does the surrogate modelling effect the performance inmulti-objective Bayesian optimisation (BO)? Multi-objective BO can be used to find a set of approximatedPareto optimal solutions in the least number of function evaluations. There are typically two approachesfor modelling in multi-objective BO: mono-surrogate or multi-surrogate. You will want to compare thesetwo approaches and propose some suggestions to use such approaches multi-objective BO. You can useone mono-surrogate and one multi-surrogate approach in this project.

3 Software tools and packages

In planning your empirical work it is worth noting there are many existing open source packages containingimplementations a number of pre-existing multi-objective optimisers, and test problems/suites as highlightedin Workshop 2.

4 Submission

The document body of the pdf format report should be no more than 4 pages (excluding references) in length.It should be typeset in LATEX, using the style file provided on the COMM510 ELE page and be submitted by12pm (midday) on the date specified on the cover page, using the electronic BART submission system.If you are less familiar with LATEX, we suggest you take advantage of the institutional Overleaf accounthttps://www.overleaf.com, which you will benefit from if you register on the website with your universityemail address. This includes an online editor and compiler, version control, as well as ‘how to' guides fortypesetting using LATEX(as covered in workshop 2).

This IT Computer Science  Assignment has been solved by our IT Computer Science Expert at Exam Question Bank. Our Assignment Writing Experts are efficient to provide a fresh solution to this question. We are serving more than 10000+ Students in Australia, UK & US by helping them to score HD in their academics. Our Experts are well trained to follow all marking rubrics & referencing Styles. Be it a used or new solution, the quality of the work submitted by our assignment experts remains unhampered. You may continue to expect the same or even better quality with the used and new assignment solution files respectively. There’s one thing to be noticed you could choose one between the two and acquire an HD either way. You could choose a new assignment solution file to get yourself an exclusive, plagiarism (with free Turn tin file), expert quality assignment or order an old solution file that was considered worthy of the highest distinction.