Computing Science Seminars, Spring 2014
Welcome to the Computing Science Seminars. Unless otherwise stated, seminars will take place in Room 4108, Cottrell Building from 15.00 to 16.00 on Friday afternoons during semester time.
Spring 2014
| Date | Presenter | Title/Abstract |
|---|---|---|
| Friday 14 Feb |
Prof Leslie S.
Smith Computing Science and Mathematics University of Stirling |
Putting Intelligence into
Computational Intelligence. What has intelligence
got to do with computational intelligence? Is there
anything that lies between the clever programming and
bottom-up pattern recognition that makes up the
(relatively undefined) field of Artificial Intelligence,
and the layperson's interpretation of intelligence? Is
there something between consciousness and awareness, and
computational Intelligence? I will argue that context and
invariance in sensory systems are critical, and that we
need to reconsider the role of time in artificial systems,
in order to build systems that can perform usefully in
real environments |
| Wednesday 19 Feb |
Dr
Thomas Stidsen Associate Professor DTU Management Engineering Technical University of Denmark |
Timetabling at DTU and the
Generalized Meeting Planning Problem. After a brief
introduction to the Operations Research at DTU-Management
I will present a recent article about meeting planning
using Dantzig-Wolfe decomposition and applying a Branch
& Price algorithm to the problem. The approach is
tested on two separate problems: Parent Consultation
Planning and Supervisor Consultation Planning, both
problems occuring in the Danish high schools. The approach
is tested on 100 real life instances of each problem type
and compared to an Adaptive Large Neighbourhood Search
(ALNS) search meta-heuristic and to a Mixed Integer
Programming formulation. Excellent performance is
obtained. Afterwards I will give a short, more general,
presentaion of the advantages and dis-advantages of using
Dantzig-Wolfe/Branch & Price. If time permits it, I
will furthermore present some other research work in
Multi-Objective Mixed Integer Programming optimization,
using a Branch & Bound algorithm. |
| Friday 28 Feb |
No Seminar
|
|
| Friday 7 March |
Prof
Hugo Terashima-Marin Director of Graduate Programs Tecnológico de Monterrey Campus Monterrey, Mexico |
Towards the Generality of Hyper-heuristics for Solving
Bin Packing Problems. The talk will focus on
describing an evolutionary hyper-heuristic framework that
is capable of tackling various types of bin packing
problems: one-dimensional, two-dimensional regular
(rectangular) and two-dimensional irregular (concave and
convex polygons). The approach is based on a genetic
algorithm (GA) with variable-length individuals
representing selective hyper-heuristics that gradually
build solutions and are able to decide what to do next, at
each step. In selective hyper-heuristics, a high-level
heuristic controls the application of single heuristics.
Several issues are addressed in this investigation such as
characterizing the problem instances to define the
adequate state and chromosome representations, deciding on
the proper set of single heuristics, and selecting and
generating problem instances to construct the testbed.
Once a hyper-heuristic has been produced by using a
training set of instances, it is reused on an unseen
instance set. Results are comparable to those obtained by
using the best choice of single heuristic for each
problem, but without the considerable expense of
determining which single heuristic is the best for that
problem. This investigation also provides an insightful
empirical analysis and discussion on the application of
single heuristics within a hyper-heuristic, aimed at
understanding their frequency, alternations and
interactions during the solution
process.
|
Friday 14 March |
Dr Yuhang Chen Lecturer Institute of Mechanical Process and Energy Engineering Heriot-Watt University, Edinburgh |
Tissue Mechanics: from Microstructure to Function. Unlike the classical biomechanics and one of its major sub-fields cell mechanics, instead of focusing on either system or cell/sub-cellular behaviours, tissue mechanics considers interaction between them at meso-scale which bridges our understanding on human physiology at multiple length scales. One of the critical questions that remain to be resolved in tissue mechanics is ‘how cell activities ‘express’ at system level and how changes taking place at organ level ‘regulate’ the cellular behaviours’. In our group we focus on tissue microstructure as an intermediate player and attempt to answer this question using analytical and engineering approaches. In this talk, some examples that are currently being studied in our group will be presented, including enhanced clinical diagnostics for bone by nonlinear mechanics, diagnostics of cancer in soft tissue (e.g. prostate) and coronary artery disease using dynamic mechanical sensing as well as in vivo design optimisation of tissue scaffold microarchitectures. |
| Friday 21 March |
Dr
Una Benlic Postdoctoral Research Assistant Computing Science and Mathematics University of Stirling |
Breakout Local Search: Application to
Gate Allocation Problem. Breakout Local Search (BLS)
is a recent variant of Iterated Local Search with a
particular emphasis on the importance of perturbation. It
explores the search space by a joint use of a local search
procedure (usually a simple descent/ascent algorithm) and
a diversification mechanism which adaptively determines
the number and type of perturbation moves by considering
some information related to the search state. In spite of
its conceptual simplicity, BLS often shows to be highly
competitive with some well-established metaheuristics.
Moreover, it is among the current state-of-art algorithms
for several classic NP-hard combinatorial problems. This
seminar presents an application of BLS to gate allocation,
one of the most important and complex airport related
problems. |
| Friday 28 March |
Mid Semester Break
|
Mid Semester Break |
| Friday 4 April Room: 2V1 |
Prof Ken
Turner Computing Science and Mathematics University of Stirling |
50 Years of Computing and Beyond.This
talk will cover my experiences of computing during the
past 50 years. A lot has changed in this period. It is
therefore hard to know how things used to be, or how we
got to where we are today. The talk will consist of
personal recollections and anecdotes, with the aim of
filling in a bit of the history. The content will be
light-hearted and pictorial rather than heavyweight and
research-oriented. Some thoughts will be offered as to how
computing might evolve over the next 50 years. |
| Friday 11 April |
Dr Rik
Sarkar Lecturer Department of Informatics University of Edinburgh |
Distributed Submodular Maximization:
Identifying Representative Elements in Massive Data.
Selecting a small representative subset from a large
dataset is a type of optimization problem called
submodular maximization. Practical examples include: which
5 news articles or web sites will give me the most
important information of the day? Or, which places in the
city are best for putting up advertisements? Many
important computational challenges such as selecting
influential nodes in a social network, exemplar based
clustering, document summarization etc belong to this
category. With the emergence of large datasets, the
classical centralized algorithms for submodular
optimization are no longer adequate, and such problems
must be solved distributively, in parallel. In this talk I will describe a simple randomized protocol that uses map-reduce style distributed computation and can produce provably good results, with further improvements when the data carries some geometric structure. The talk is based on a paper recently published at NIPS (Neural Information Processing Systems) 2013.
|
| Friday 18 April |
Good Friday |
Good Friday
|
| Friday 25 April |
Dr Paul
Fergus Research Fellow Networked Appliances Laboratory School of Computing and Mathematical Sciences Liverpool John Moores University |
TBA
|
| Friday 2 May |
Prof
Emma Hart Director Centre of Emergent Computing Institute for Informatics and Digital Innovation Edinburgh Napier University |
A Immune Inspired lifelong Learning
Method for Solving Combinatorial Optimisation Problems.
The previous two decades have seen significant advances in
meta-heuristic and hyper-heuristic optimisation techniques
that are able to quickly find optimal or near-optimal
solutions to problem instances in many combinatorial
optimisation domains. Despite many successful applications
of both these approaches, some common weaknesses exist in
that if the nature of the problems to be solved changes
over time, then algorithms needs to be periodically
re-tuned. Furthermore, many approaches are likely to be
inefficient, starting from a clean slate every time
a problem is solved, therefore failing To exploit
previously learned knowledge. In contrast, in the field of machine-learning, a number of recent proposals suggest that learning algorithms should exhibit life-long learning, retaining knowledge and using it to improve learning in the future. Looking to nature, we observe that the natural immune system exhibits many properties of a life-long learning system that could be computationally exploited. I will give a brief overview of the immune system, focusing on highlighting its relevant computational properties and then show how it can be used to construct a lifelong learning optimisation system. The system is shown to adapt to new problems, exhibit memory, and produce efficient and effective solutions when tested on a large corpus of bin-packing problems. |
| Friday 9 May |
Dr
Jun Chen School of Engineering University of Lincoln |
Towards a More Cost Effective and
Environmentally Friendly Airport Surface Movement
through Active Routing and Guidance. The Airport
Surface Movement problem is one of the challenging
real-world optimisation problems found at airports. In
this talk, a new concept called Active Routing and
Guidance is proposed and set into a wider context of
previous research which saw an evolution from simple
routing to a complex decision support system for both air
traffic controllers and pilots. In its nature, the airport surface movement problem is a multi-objective multi-component optimisation problem which combines two components: scheduling and routing of aircraft and speed profile optimisation. In this talk, a multi-objective optimisation framework is presented which combines lower-level algorithms for each component. Different speed up techniques are introduced including a heuristic for the speed profile optimisation and an adaptive speed profile database. The performance of the proposed algorithms is demonstrated using real data from major airports. Finally, a metaheuristic approach is introduced to integrate the airport surface movement problem with other airport problems. |
Top image: This is an
illustrated example of running the Epsilon-constraint
algorithm in order to maximise two objectives: find an optimal
solution for objective 1; restrict the solution space according
to the solution's value for objective 2 and look for an optimum
solution of objective 1 in that space; repeat the previous step
until there are no more solutions to be found. Any dominated
solutions need to be filtered out of the set of solutions.
Courtesy of Dr.
Nadarajen Veerapen
Last Updated: 02 May 2014.