IMAT5232 Computational Intelligence Optimisation

Computational Intelligence Optimisation (CIO) is a subject that integrates artificial intelligence into algorithms for solving optimisation problems that could not be solved by exact methods. Thus, CIO is the subject that defines and designs algorithms such as metaheuristics for optimisation, i.e. general purpose algorithms. This makes CIO the subject that tackles optimisation problems in engineering, economics, and applied science. This subject contains algorithmic structure based on metaphors such as evolution and collective intelligence. This module will provide students with an appreciation of both, theoretical and implementation issues of CIO algorithms. Selected algorithms (negotiated between the lecturer and each student) will be studied in practical work.

Outline content:

• Generalities: Definition and postulate of optimisation problems, fitness landscape and problem features, No Free Lunch Theorem
• Brief introduction to exact optimisation via differentiation
• Classical derivative free methods: General concepts of Rosenbrock, Hooke-Jeeves, Nelder-Mead Algorithms, Simulated Annealing, Multi-start search
• Popular population-based algorithms: Genetic Algorithms, Evolution Strategy, other examples of modern evolutionary approaches, Particle Swarm Optimisation
• Differential Evolution, other examples of modern swarm intelligence algorithms, perturbation mechanisms of an algorithmic scheme
• An overview on modern approaches: Adaptive Systems, Hyper-heuristics, Memetic Computing
• An overview on special problems: Multi-objective, noisy, dynamic, computationally expensive, and large scale problems
• Application examples: control theory, image processing

Learning outcomes. On successful completion of this module a student will be able to:

1. Demonstrate a comprehensive understanding of Computational Intelligence Optimisation;
2. Be able to critically apply and implement the taught CIO algorithms to given test problems.

Recommended Texts

Algorithmic issues in computational intelligence optimization : from design to implementation, from implementation to design. F. Caraffini (download)

Introduction to Evolutionary Computing A. E. Eiben, J. Smith by Springer

Introduction to Genetic Algorithms Sivanandam, S.N.; Deepa, S.N. Published By: Springer

IMAT5235 Artificial Neural Networks

This module provides a detailed appraisal of several aspects of neural network computing. It provides a history of the subject and then covers in detail the various network paradigms which have become established as useful computational tools. Applications will be discussed and students will be introduced to problem domains where problem instances may be amenable to solution by neural network techniques. Whilst the module will concentrate on an Engineering approach there will also be discussion of the use of networks for cognitive modelling.

Topics include:

• Historical account
• Learning paradigms
• Feed Forward Networks (Classical and Modern approaches)
• Self-organising maps
• Recurrent Networks
• Applications to: pattern recognition; classification problems; data modelling; time series; cognitive modelling

Learning outcomes. On successful completion of this module a student will be able to:

1. Apply modelling approaches which use neural networks to solve computational problems;
2. Implement a variety of network solutions;
3. Have a comprehensive knowledge of the successful application of neural networks to several problem domains and be capable of judging whether the neural computational approach might be fruitful in a novel situation;
4. Participate in the peer review process.

Recommended Text

Steeb, 2011, The Non-Linear Workbook, available on Amazon.

IMAT5234 Applied Computational Intelligence

The purpose of this module is to enable students to appreciate the historical, philosophical and future implications of AI in relation to both theoretical and practical aspects and to investigate at least one application area in depth.

Topics include:

• History & philosopshy of AI: history of developments in AI; exploration of prevalent philosophical views and theories e.g. Searle, Dennett, Penrose;
• Expert Systems: Knowledge acquisition, representation and search, expert system shells.
• Applications: an exploration of applications in a variety of different areas will be achieved by combinations of study of current research papers, guest speakers, tutors’ own research & the investigative work of the students within the module.

Learning outcomes. On successful completion of this module a student will be able to:

1. Apply AI techniques to given practical problems
2. Recognise the multi-disciplinary nature of AI and its potential application areas.
3. Critically appraise relevant literature in order to formulate a plan for their own practical/experimental work
4. Synthesise a solution to a problem (planned in LO3) and evaluate the solution

Recommended Text

Intelligent Systems for Engineers & Scientists, Hopgood, CRC Press, 2001 (the newer edition may be available now)

IMAT5233 Intelligent Mobile Robots (MSc ISR)

This module builds on the material covered in Mobile Robots to provide a comprehensive understanding of autonomous mobile robots and autonomous navigation. The aim of the course is to enable the student to comprehend and argue constructively the space, reasoning and navigation. In this module students will be required to analyse, evaluate and construct odometry systems, maps, navigation plans and localisation techniques for mobile robots. Issues related to the sensing, representing and modelling of the environment will be assessed. Some algorithmic solutions will be synthesised and assessed. Advanced issues such as simultaneous localisation and mapping will be critically discussed.

Topics include:

• Representational issues and reasoning representing space, the robot and action representing knowledge and perception path-planning
• Localisation
• Mapping unknown environments
• Handling sensory uncertainties using stochastic methods
• Applying particle filters to localisation
• Having some understanding of how Kalman filters can be applied to the problem of simultaneous localisation and mapping

Learning outcomes. On successful completion of this module a student will be able to:

1. Demonstrate a comprehensive understanding of the various approaches to incorporating spacial awareness and navigation in mobile robot behaviour.
2. Demonstrate the ability to critically evaluate navigation and localisation solutions.

Recommended Text

Murphy, 2000, An Introduction to AI Robotics, MIT Press

IMAT5238-2017-2 Data Mining, Techniques and Application (MSc IS)

Data is collected and stored in all different types of organisations - commercial, governmental, educational. Every day hundreds of terabytes of data are circulated via the Internet. Organisations are now making the most out of the data they electronically capture by looking for hidden patterns and extracting meaningful information for decision making purposes. For example, in marketing, customers who are most likely to buy certain products and services should be targeted. In fraud detection, it is of interest to investigate unusual behaviour patterns to identify insurance claims, cellular phone calls and credit card purchases that are most likely to be fraudulent. Data mining is a collection of tools, methods and statistical techniques for exploring and modelling relationships in large amounts of data, to enable meaningful information to be extracted for decision making purposes. The aim of this module is to review the data mining methods and techniques available for uncovering important information from large data sets and to know when and how to use a particular technique effectively. The module will enable the student to develop an in-depth knowledge of applying data mining methods and techniques and interpreting the statistical results in relevant problem domains. Current application areas and research topics in data mining will also be discussed and students will be expected to contribute to these discussions to increase their background knowledge and understanding of issues and developments associated with data mining. The module uses the data mining tool SAS Enterprise Miner.

Topics include:

• Introduction to data mining
• Data mining methodology
• Exploratory data analysis including association analysis
• Cluster analysis in data mining
• Predictive modelling using Regression
• Predictive modelling using Decision Trees
• Predictive modelling using Neural Networks
• Model evaluation: Comparing Candidate Models
• Model implementation: Generating and Using Score Code
• Current research and application in data mining

Learning outcomes. On successful completion of this module a student will be able to:

1. Have an in-depth understanding of the key concepts of data mining
2. Appreciate the breadth of areas of application and research in data mining
3. Systematically apply appropriate data mining methods and techniques for particular problem domains
4. Correctly interpret and critically evaluate the results to make informed decisions

Recommended Texts

Berry Michael and Linoff Gordon (2011). Data Mining Techniques. 3rd ed. ISBN 978-0-470-65093-6 (Wiley)

SAS Publishing (2003): Data Mining using SAS Enterprise Miner: A Case Study Approach. 2nd ed. ISBN 1-59047-190-3 (SAS Publishing.)

Baker Stephen (2009). They've got your number: Data, Digits and Destiny - how the Numerati are changing our lives.