Computer Science Home  

2011-2012

Research@NoU-CS          
The LANCS Initiative
ASAP Research Group

Ender Özcan
Office: C86
T:+44(0) 115 95 15544
F:+44(0) 115 9514254

exo At cs-nott-ac-uk (replace all - with dot)


Home Research Publications Activities Teaching

G52GRP: Software Engineering Group Project

For details, such as reading material, deadlines, assessment, and more, please refer to the G52GRP module web page.

AUDRI: An Automated Driver - The Student Becomes The Master

Apprenticeship learning also known as learning by demonstration or imitation learning, is a machine learning approach for generalising the demonstrations provided by an expert. This project involves implementation of a Java based tool which contains a simulator for driving a car on a high way with multiple lanes and a chosen apprenticeship learning system to learn how to drive safely avoiding collisions. The car under control will have a fixed speed higher than the other cars on the way. The left lane should be preferred over the middle lane over the right lane, over driving off-road. The goal is to generate an AUtomated DRIver (AUDRI) by learning to drive properly on the highway mimicking the behaviour of a demonstrator and if possible, even improving the learning process by using the data obtained from some previous demonstrators. The tool should also allow a user to evaluate/observe his/her safe driving skills as compared to AUDRI on the same simulation interface.

Note that Weka provides a rich set of techniques that can be used for apprenticeship learning. The focus of this project is a software-engineering effort to design the overall framework to bring the components together; user interface, driving simulator and Weka, and choose an appropriate apprenticeship learning technique.

Resources:

1) Weka: http://en.wikipedia.org/wiki/Weka_(machine_learning)

2) Abbeel, P., Ng, A.Y.: Apprenticeship learning via inverse reinforcement learning. In: Proceedings of the twenty-first international conference on Machine learning. pp. 1-8. ICML '04, ACM, New York, NY, USA (2004)

A SAMPLE FROM PREVIOUS GROUP PROJECTS

A Selection Hyper-heuristic Visualisation Tool for HyFlex v1.0

Group Members:

Name
Cai Leran
Ellis Jacob
Hood Steven
Howard Thomas
Mukherjee Shayoke
Wong Hon Wai

Description

A hyper-heuristic is ?a search method or learning mechanism for selecting or generating heuristics? to solve computationally difficult problems. There is a growing interest in hyper-heuristics representing a class of methodologies with the common goal of automating the design and tuning of heuristic methods in search and optimisation. The focus of this study is selection hyper-heuristics. Under such a single point based search framework, a given complete solution is improved iteratively using a set of low level heuristics. At each step, a solution passes through two successive stages: heuristic selection and move acceptance. The heuristic selection mechanism chooses and applies a low level heuristic to a candidate solution producing a new complete solution. Then, the acceptance strategy decides whether to continue the search process using the new solution or the one in hand. This process continues until a termination criterion is satisfied and then the best solution found so far is returned. This project involves in designing and implementing a Java applet (or application) that illustrates how selection hyper-heuristics (particularly, choice function, reinforcement learning, simple random combined with accept all moves, improving and equal) work and provides relevant statistics on a given optimisation problem (problem domain to be decided: Bin packing, Max-SAT, flow shop, nurse rostering, TSP or VRP) under HyFlex v1.01, a Java software library implemented for rapid development of hyper-heuristics and research.
Initial reading (with a focus on selection hyper-heuristics):

  • E. K. Burke, M. Gendreau, M. Hyde, G. Kendall, G. Ochoa, E. ?zcan, R. Qu, Hyper-heuristics: A Survey of the State of tdariohe Art, to appear in the Journal of the Operational Research Society, DOI:10.1057/jors.2013.71.

  • J. H. Drake, E. Ozcan and E. K. Burke, An Improved Choice Function Heuristic Selection for Cross Domain Heuristic Search, PPSN 2012, Lecture Notes in Computer Science 7492, part II, pp. 307-316, 2012.

  • G. Ochoa, M. Hyde, T. Curtois, J. A. Vazquez-Rodriguez, J. Walker, M. Gendreau, G. Kendall, B. McCollum, A. J. Parkes, S. Petrovic, E. K. Burke, EvoCOP 2012, LNCS 7245, pp. 136-147, 2012.

  • E. ?zcan, M. Misir , G. Ochoa, E. K. Burke, A Reinforcement Learning - Great-Deluge Hyper-heuristic for Examination Timetabling, International Journal of Applied Metaheuristic Computing, 1(1), pp. 39-59, 2010.


An Intelligent User Interface for Recognising Free-hand Graph Drawings

Group members:

Name
Cox, Robin John Dashwood
Hao, Yajuan
Leask, Samuel
Mandalia, Kevin
Sansom, Adam
Wang, Ruoyang

Description:

Free-hand drawings is a natural mode of interaction used in a variety of environments. As the use of pen-based systems such as tablet computers and smart-boards grow, sketch recognition has regained attention enabling natural pen-based interfaces.

This project involves in creating a natural and easy-to-use intelligent user interface by combining techniques from computer vision, machine learning, computer graphics and human-computer interfaces. Interactive multimedia such as computer simulations and animations received increased attention over the years as supplementary teaching tools and have now become integral components of most engineering and science curriculums. One way to increase the utility of such simulations and animations is to make them easier to use. In this project, a pen-based multimodal interface will be implemented based on sketch recognition algorithms and a speech recognition system (Windows SAPI). The goal of the project is to recognize spoken commands and interpret free-hand drawings for constructing weighted and unweighted graphs which functions as a front-end to a shortest path and minimum spanning tree (MST) algorithm simulators.

Note that the various subcomponents already exist: the focus of this project is a software-engineering effort to design the overall framework to bring the components together, the design of the user interface, and animation for two graph algorithms.

Initial reading:

  • H. Dibeklioglu, T. M. Sezgin, E. Ozcan. A Recognizer for Free-Hand Graph Drawings. International Workshop on Pen-Based Learning Technologies, Catania, Italy, May 24-25 (2007).
  • T. M. Sezgin, R. Davis: Sketch recognition in interspersed drawings using time-based graphical models. Computers & Graphics (CG) 32(5):500-510 (2008)
  • T. M. Sezgin, I. Davies, P. Robinson: Multimodal inference for driver-vehicle interaction. ICMI 2009:193-198
  • A. Blessing, T. M. Sezgin, R. Arandjelovic, P. Robinson, A multimodal interface for road design. Workshop on Sketch Recognition, International Conference on Intelligent User Interfaces, Sanibel, FL, February (2009).

A Multimodal Interface for Free-Hand Graph Drawings

Group members:

Name
Bautista, Kenneth Jose
Bourne, Thomas Alexander
Hassoun, Tarek
Li, Jie
Slee-Egeler, Tariq
Wang, Xiao

Description:

This project involves in creating a natural and easy to use intelligent user interface by combining techniques from computer vision, machine learning, computer graphics and human-computer interfaces. Interactive multimedia such as computer simulations and animations received increased attention over the years as supplementary teaching tools and have now become integral components of most engineering and science curriculums. One way to boost the utility of such simulations and animations is to make them easier to use. In this project, a pen-based intelligent interface will be implemented and combined with a speech recognition system (Windows SAPI). The goal is to recognize spoken commands and interpret free-hand drawings for constructing weighted and unweighted graphs which functions as a front-end to a shortest path and minimum spanning tree (MST) algorithm simulators. The usability of the different types of interfaces (such as, a WIMP based interface and hybrids) will be evaluated.

Initial Reading:

  • H. Dibeklioglu, T. M. Sezgin, E. Ozcan. A Recognizer for Free-Hand Graph Drawings. International Workshop on Pen-Based Learning Technologies, Catania, Italy, May 24-25 (2007).
  • A. Blessing, T. M. Sezgin, R. Arandjelovic, P. Robinson. A multimodal interface for road design. Workshop on Sketch Recognition, International Conference on Intelligent User Interfaces, Sanibel, FL, February (2009).

Visualisation Tool for a Choice Hyper-Heuristic

Group members:

Name
Barton, Thomas
Jenkinson, Ben
Jermstad, Alexander Shevlin
Lao, Jingqi
Luland, Steven
Zhang, Chao

Description:

Hyper-heuristics can be defined as "methodologies to choose heuristics". There is a growing interest in hyper-heuristics as powerful tools in search and optimisation. A randomly generated initial candidate solution is improved iteratively using a set of low level heuristics in a simple choice hyper-heuristic framework. At each iteration, a given solution passes through two successive stages: heuristic selection and acceptance. The heuristic selection mechanism chooses and applies a low level heuristic to a candidate solution producing a new solution. Then, the acceptance strategy decides whether to continue the search process using the new solution or the one at hand. This project involves in designing and implementing a Java applet (or application) that demonstrates how a choice hyper-heuristic works on an optimisation problem that requires binary representation.