University Staff Teaching Allocation: Formulating and Optimising a Many-Objective Problem

Add to your list(s) Download to your calendar using vCal

• Prof Jonathan Fieldsend (Department of Computer Science, University of Exeter)
• Monday 10 July 2017, 11:00-12:00
• Computer Science, The Sloman Lounge (UG).

If you have a question about this talk, please contact Hector Basevi.

Host: Prof Joshua Knowles

Webpage: http://emps.exeter.ac.uk/computer-science/staff/jefields

Abstract: The allocation of university staff to teaching (often decided by heads of department in conjunction with directors of education, in consultation with the staff concerned) often exhibits a range competing objectives. These can include fair distribution of load between staff, minimising peak load, minimising load imbalance across terms, minimising staff dissatisfaction with their allocation, minimising churn in teaching allocation between years, and minimising student dissatisfaction.

This talk illustrates the use of an augmented version of a popular many-objective optimiser (NSGA-III) to undertake the seven-objective optimisation of the teaching staff allocation problem for the University of Exeter Computer Science department. This tool has been used interactively to propose the teaching allocation for the academic year 2017/18, and we also compare results to a ‘hand-tuned’ allocation on the current year’s data (2016/17).

The talk will show how we can derive mathematical forms for the various allocation quality criteria, and also how we can determine the maximum possible values for all criteria identified and the minimum values for most exactly (with lower bounds on the remaining criteria). For many of the optimisation criteria, when considered in isolation, an optimal solution may also be obtained rapidly. We demonstrate the advantage of utilising such extreme solutions to drastically improve the optimisation efficiency in this particular many-objective optimisation problem. We also identify issues that many-objective decomposition-based optimisers can experience due to selection between generations. The talk will conclude with proposed advances and additional functionality we intend to develop for future years, given our experience with the developed tool this year. Issues regarding allocation robustness are also discussed.

This talk is part of the Artificial Intelligence and Natural Computation seminars series.

This talk is included in these lists:

• Artificial Intelligence and Natural Computation seminars
• Computer Science Departmental Series
• Computer Science Distinguished Seminars
• Computer Science, The Sloman Lounge (UG)
• computer sience

Note that ex-directory lists are not shown.