University of St Andrews

Key information

SCOTCAT credits : 15

ECTS credits : 7

Level : SCQF level 11

Semester: Both

Availability restrictions: Available only to students studying the PG Cert/PG Dip/MSc in Data Science (Digital)

Many problems in modern Data Science, such as planning and scheduling, involve solutions with integer values and options for solutions that grow in size very rapidly. This module covers the theory, tools and technologies developed and used to solve problems in Integer Programming and Combinatorial Optimization. Data science problems often require a solution from a discrete set of values. Examples include scheduling, resource allocation, and path selection. For many important problems, a numeric optimization formulation can be useful for finding optimal or near-optimal solutions. In other cases, the generality provided by discrete optimization tools and techniques may be more efficient in the identification of high-quality solutions and may also supply useful additional information that can be leveraged to solve future problems. For CS5929 the types of optimisation include greedy methods; local search (Hill Climbing, Simulated Annealing); linear programming; constraint programming; and hybrid methods (LNS).

Learning and teaching methods and delivery

Weekly contact: Students should expect to engage in approximately six tutorials over the course of the module, which will be scheduled with an awareness of the pace at which they are progressing, rather than at a fixed time each week. Students should consider the amount of independent study time this module involves when planning their learning.

Scheduled learning hours: 6

Guided independent study hours: 148

Assessment pattern

As used by St Andrews: Coursework = 100%

As defined by QAA
Written examinations : 0%
Practical examinations : 0%
Coursework: 0%

Re-assessment: Coursework = 100%

Personnel

Module coordinator: Professor T W Kelsey
Module teaching staff: Dr Joan Espasa Arxer and Dr Ozgur Akgun

Intended learning outcomes

• Understand core concepts in the field of Discrete Optimisation.
• Be able to use declarative modelling for optimisation.
• Be able to use candidate approaches for solving these problems, including Constraint Programming, Boolean Satisfiability (SAT), SAT Modulo Theories (SMT).
• Understand how discrete optimisation can be applied together with machine learning.
• Understand how a particular alternative (model, solver, and configuration) can be robustly chosen for a given problem.