Category Archives: Events

Ensembles Methods for Optimisation Workshop

gecco_logo_HQ

Ensemble Methods for Optimisation

Workshop @ GECCO 2017, Berlin

Organisers: Emma Hart and Kevin Sim

In the field of machine-learning, ensemble-methods that combine decisions from multiple learning algorithms to make accurate predictions have been the focus of intense research for well over a decade, with a broad range of empirical results underpinned by a sound theoretical understanding. Ensemble methods have also found favour within the constraint satisfaction and satisfiability domains where they are commonly referred to as portfolio methods. In the latter case, portfolios tend to be composed from exact solvers, and are evaluated according to run-time metrics.

On the other hand, research in ensemble-methods using meta-heuristic algorithms – in which solution quality rather than run-time is the driving factor – lags behind machine-learning and satisfiability research in both theory and practice. Many fundamental questions remain with respect to how to construct, and design ensembles that will be addressed during the workshop:

  • How should we select algorithms to form an ensemble?
  • How large should the selection pool be – and where do we find algorithms to form the pool?
  • Are automated algorithm generation techniques required to design new algorithms to provide a large enough pool?
  • Machine-learning theory suggests that diversity between components is a key factor – what diversity measures can be used to successfully distinguish between meta-heuristic algorithms?
  • How should the ensemble operate? Algorithms might collaborate, i.e. the computational budget is divided between algorithms within the ensemble, or cooperate, in that different algorithms solve different instances?
  • What domains are ensemble methods best suited to?
Submissions
We invite paper submissions describing technical work as well as conceptual/visionary short papers. Demos accompanied by short papers are also welcomed. The workshop encourages presentation of work in early stages in order to encourage discussion amongst participants.
Technical papers have a limit of 8 pages.
Conceptual and Demo papers have a limit of 2 pages.
Submissions should be emailed to e DOT hart AT napier.ac.uk with the subject heading:
EfO Submission
All papers should be in ACM (Association for Computing Machinery)
format. Please see GECCO 2017 http://gecco-2017.sigevo.org for
details. Papers should not be anonymised. All papers should be
submitted in PDF format.
All accepted papers will be presented at GI-2017 and will appear in
the GECCO workshop volume.
Key Dates
Paper submission deadline:  March 29, 2017 (no extensions permitted)

Workshop on the Mathematical Modelling of the Risk of Wind Damage to Forests

Background

Wind is a major disturbance agent in forests and a key part of the dynamics of many forest ecosystems and although this has long been recognised in boreal and temperate forests there is now increasing evidence of its importance in tropical forests. In addition, the high levels of damage that can occur during storms have important economic and social impacts, particularly for managed forests. For example, in European forests wind is responsible for more than 50 % of all damage by volume and the cost of such damage can be very high (e.g. € 6 billion in France from storms Lothar and Martin in 1999, and €2.4 billion in Sweden after storm Gudrun in 2005). There is clear evidence that damage levels to forests have been increasing over the past century due to a mixture of changing climate and forest management practice, and damage levels are predicted to continue to increase. To understand the influence of wind damage on forest ecosystems or to manage forests in order to mitigate the risk of wind damage, it is necessary to have methods for predicting the levels of damage at different wind speeds and the risk of such wind speed occurrence.

Normally mathematical models are used to make assessments of wind damage risk to forests and such modelling has been active for approximately 15 years following two main approaches:

  1. 1:  Statistical models developed from observations of damage in well monitored forests with information on soil, tree size, elevation, forest management, etc.
  2. 2:  Mechanistic models that use methods from civil/mechanical engineering to directly predict the wind loading on trees and the wind speed at which they will fail.

Statistical models generally make good predictions within the locale for which they were developed but do not always translate well to other locations or conditions. Mechanistic models are usually not as accurate as statistical models for a specific area but are easier to transfer to other situations. They also the possibility of providing wind risk evaluations at all scales from individual trees to forest, regional, national and potentially global scales. However, it is becoming clear from the literature and conference discussions (e.g. IUFRO Wind and Trees Conference, August 2014, Brazil) that current mechanistic models have reached a limit in their capability (spatially and temporarily) and a re-evaluation of the approach is required to make progress. Therefore it was recommended by Prof. Steve Mitchell, chairman of the IUFRO working group 8.03.06, that a workshop focussed on mechanistic wind risk modelling would be of value in starting the discussion on the best ways to develop such models in the future.

Goals

The planned outputs of the workshop are:

  1. A report providing recommendations for the future of forest wind damage modelling, with a list of available tools/model and data, together with a discussion of the most appropriate methodologies for different temporal and spatial scales. The report will also contain an assessment of barriers to the development of new mechanistic modelling approaches and the major gaps in our knowledge.
  2. A review article on methods for modelling forest wind risk, from the single tree to national scale, for submission to Environmental Modelling and Software

Overview

The workshop was organised by ROLL collaborator Prof. Barry Gardiner and took place in Arcachon (near Bordeaux, France) from 28-30th October and brought together forest wind risk modellers from around the world along with experts in subjects required to improve the models (e.g. meso-scale airflow modellers, wind climate scientists, risk analysis engineers, etc.), but who have to date not always been actively involved in forest wind damage risk modelling.  Professor Emma Hart gave a talk on the opportunities for using Machine Learning techniques within the industry, and posed the question of whether novel optimisation methods might be developed given that potential solutions can be easily evaluated by replacing expensive simulations with tools like neural networks that approximate solutions. Slides from her talk are available here : WindDamageForests

(photograph courtesy of Barry Gardiner)

AHDB Smart Agriculture 2015

The exchange of innovation and technology is essential for any industry to overcome current issues and adapt to face future challenges and opportunities. The use of technology in agriculture & horticulture is nothing new. Farmers and growers are continuously seeking ways to reduce cost, improve quality and advance productivity. But what can be learned from other industries? What is out there and how can we exploit advances in engineering, manufacturing, computer science or the medical arena and apply them to the future of precision farming?

AHDB’s Smart Agriculture Conference was an opportunity for scientists, researchers and engineers across multiple disciplines to build relationships that stimulate discussion and identify potential research and application opportunities that address bridge the gap between farmer need and technological development. The conference focused on smart technology encompassing engineering, sensors, nano technology, decision support systems and robotics .

AHDBlogo

Professor Hart gave a presentation focusing on the opportunities and benefits of applying optimisation techniques with the Agricultural Industry, highlighting potential applications within scheduling, decision making and parameter optimisation. A link to the slides and a video of the presentation are available.

Workshop on Real-World Problems@GECCO 2014

The 2nd workshop on Real-World Problem and Problem Understanding was held at GECCO 2014 in Vancouver. The workshop offered a mixed programme of invited talks,  peer-reviewed paper presentations and a demo paper, focusing on various aspects of tackling real-world problems with meta-heuristic algorithms.

The workshop opened with a talk by Prof. Darrell Whitley from Colorado State University, drawing on his many years of experience in the field to discuss Why Are There Not More Applications of Evolutionary Algorithms? Darrel highlighted a number of factors, highlighting in the main the gap between the types of problems typically solved by academics and the complex, constrained nature of real-world problems.

This was followed by two talks from authors of accepted papers:

  • A Comparison of Antenna Placement Algorithms, Abhinav Jauhri, Jason D. Lohn, Derek S. Linden
  • Hyper-Heuristic Genetic Algorithm for Solving Frequency Assignment Problem in TD-SCDMA , Chao Yang, Shuming Peng, Bin Jiang, Lei Wang, Renfa Li

 Dr. Anna Esparcia-Alcazar, head of R&D at S2-Grupo provided an entertaining view from an industry perspective, illustrating her talk with a variety of anecdotes and advice gleaned from working with a company providing solutions to optimisation problems to SMEs. Her talk Six Impossible Things Before Breakfast: Computational Intelligence in the SME illustrated some of the ups and downs of working with a company, not all of them in the control of the meta-heuristic designer!

The workshop concluded with a demo and talk by Prof. Emma Hart, describing work recently undertaken with a company to develop an optimisation algorithm aimed at employee scheduling and routing, i.e. allocating jobs to workers and additionally routing them around the job, accounting for service-completition deadlines and maximising efficiency. An interactive GUI that enable the fitness function to be altered during the course of a run was demonstrated during the talk.

All workshop papers are available in the GECCO companion proceedings.

 

CFP: GECCO 2014 Workshop on Problem Understanding and Real-world Optimisation (PURO)

Full details here: http://www.soc.napier.ac.uk/~cs378/PuroGECCO2014/index.html

to be held as part of the
2014 Genetic and Evolutionary Computation Conference (GECCO-2014)
Vancouver, BC, Canada, July 12-16, 2014

Submission Deadline : March 28th, 2014
See http://www.soc.napier.ac.uk/~cs378/PuroGECCO2014/index.html for submission details.
Accepted workshop papers are published in their own volume by ACM.

Description

Building on the success of previous Understand Problems and Real-World Optimisation workshops, this workshop aims to provide a single forum for the presentation and discussion of works focused on optimisation problems rather than the methods for solving them. The workshop brings together the study of real-world optimisation problems with the theoretical analysis and synthesis of problems.

Workshop Scope

The workshop will focus on, but not be limited to, topics such as:

    • Methods for identifying and constructing models for new optimisation problems
    • Study of real-world optimisation problems and case studies
    • Theoretical and practical analysis of optimisation problems
    • Fitness landscape analysis of real-world problems
    • Classification and ontological analysis of problems
    • Development of and technologies for building benchmark test problems
    • Technologies and theoretical works supporting the implementation, examination and

 

Workshop Format

The workshop will run for half a day (2 sessions). We aim to hold a mixed workshop of invited talks (including at least one from industry) and paper presentations, ending with an open discussion. Authors will be invited to submit either an extended abstract or full paper, with more presentation time allocated to full paper submissions. Abstracts will also be invited from researchers wishing to provide demonstrations of software that address any of the themes of this workshop. To facilitate discussion of new ideas and approaches, authors will be encouraged to submit position papers which present new ideas for discussion or are of a conceptual nature. Papers will be accepted or rejected based on blind peer review conducted by our organising committee.

Paper/Poster Submission

Papers should be formatted using the standard ACM templates and not exceed 8 pages. Formatting instructions can be found at http://www.sigevo.org/gecco-2014/papers.html Papers should be submitted by email to k(dot)sim(at)napier.ac.uk with the subject line GECCO WORKSHOP SUBMISSION no later than March 28 2014

Draft Schedule

Session 1
Invited Talk (industry)
Workshop papers
Demos

Session 2
Invited Talk (academia)
Workshop Papers
Demos

Workshop Chairs

Kent McClymont is an associate research fellow at the University of Exeter. His research is focused on the study of multi-objective hyper-heuristic methods for solving hard real-world optimisation problems with heterogeneous encodings and novel methods for evaluating heuristics through new test problems and methodologies. He has run two previous GECCO workshops on “Understanding Problems” and is a member of the AISB committee and oversees the AISB workshop series.

Kevin Sim is a research fellow at Edinburgh Napier University, Scotland, UK. He currently works on a large national EPSRC project (EP/J021628/1) entitled “Real World Optimisation with Life-Long Learning”. His interests lie in the field of hyper-heuristics and classification algorithms. He has previously co-chaired workshops on the subject of real world optimisation problems at GECCO and EvoStar.

Gabriela Ochoa is a Lecturer at the University of Stirling in Scotland. Her research interests lie in the foundations and application of evolutionary algorithms and heuristic search methods, with emphasis on autonomous (self-*) search and fitness landscape analysis. She has published over 60 international peer reviewed papers. She is associate editor of Evolutionary Computation (MIT Press), was involved in founding the Self-* Search track at GECCO, and has organised several workshops and special sessions at international conferences.

Ed Keedwell is a Senior Lecturer in Computer Science at the University of Exeter. His research is focused on Nature-Inspired Computation techniques and their application to real-world optimisation problems in engineering and bioinformatics. He has published over 70 papers in this field and currently leads a group of 8-10 postgraduate students and postdoctoral researchers. His currently funded EPSRC research includes ‘SEQAH’ (EP/K000519/1), a project investigating the interaction between selective hyper-heuristics, heuristics and problems over time.

Important Dates

Paper/Poster Submission: March 28 2014
Notification to Authors: April 15 2014
Camera ready Submission: April 25 2014
Conference Dates: July 12-16, 2014
Workshop Dates: T.B.A.

Further Information

Workshop Website: http://www.soc.napier.ac.uk/~cs378/PuroGECCO2014/index.html
Conference website: http://www.sigevo.org/gecco-2014/

Roll Team to exhibit at SICSA Demofest

The ROLL team will be demonstrating their new software nicknamed NELLI at the SICSA Demofest in Glasgow on Tuesday 5th November.

Our goal  is to develop an automated system for solving real optimisation problems that, like humans, continuously learns over time and improve its performance with experience. The resulting Network for Life Long Learning (NELLI) produces fast, high-quality solutions and adapts quickly to changing circumstances and problem characteristics. The efficient design of the system enables NELLI to cope with the complex constraints that characterise many real-world problem.

Key features of NELLI include:

  • Automatically generated heuristics collaborate to optimise overall performance
  • Using prior knowledge to quickly provide good quality solutions while continually adapting to new unseen problems
  • Maintaining a memory that enables rapid production of solutionsNELLI

 

 

 

 

 

 

GECCO workshop: list of invited papers and accepted talks now available

Workshop on Problem Understanding and Real-World Optimisation

Saturday 6th July 2013

GECCO 2013

DOWNLOAD WORKSHOP PROGRAMME

This workshop sees the joining of two related workshops to reflect the link between stronger methods in problem understanding and the significant importance of scientific approaches to tackling real-world optimisation problems. This workshop will address the wider topic of scientific methods for analysing and solving difficult and real problems currently faced by optimisation researchers.

The workshop will continue the themes of last year’s GECCO-UP workshop and aims to provide a forum for the discussion and exploration of methods for the analysis, creation and synthesis of optimisation problems through both theoretical and experimental methods. A crucial aspect of this research is developing problems that reflect the complexity and idiosyncrasies of real-world optimisation problems. Therefore, the workshop will open with a session focused on real-world optimisation problems, in which speakers from both industry and academia discuss their experiences of what makes real-world problems difficult. The workshop as a whole provides an opportunity for discussion and interactions that might move towards bridging the gap between those problems addressed by scientific research and the types of problem addressed by commercial optimisation software vendors.

Furthermore, the continued the ethos of open debate of the previous workshops will be upheld in a discussion session held at the end of the workshop which will give an opportunity for participants to discuss and present recently published research and position papers.

Programme

Session 1
08.30-08.35   Welcome & Introduction
08.40 – 9.30 Invited Speaker: Tim Pigden, Optrak – Vehicle Routing Software: Missing from the model: Orders, shifts and the rush-hour and why they matter
09.30-09.55 The Challenges of Applying Optimization Methodology in Industry. Bogan Filipic and Tea Tusar
0-9.55-10.20 Using Graphical Information Systems to improve vehicle routing problem instances. Neil Urquhart; Catherine Scott; Emma Hart

COFFEE BREAK

Session 2
10.40-11.30   Invited Speaker: Gabriela Ochoa, University of Stirling: Search Methodologies in Real-world Software Engineering
11.35-12.00   Liger – An Open Source Integrated Optimization Environment. Ioannis Giagkiozis; Robert J. Lygoe; Peter J. Fleming
12.00-12.30   Discussion: Bridging the Gap between Academia and Industry

LUNCH

Session 3
14.00-14:45   Invited Speaker: Katherine Malan, University of Pretoria – Techniques for characterising fitness landscapes and some possible ways forward.
14.50—15.15   Recent Advances in Problem Understanding: Changes in the Landscape a Year On. Kent McClymont, University of Exeter
15.20 -15.45   EA-based Parameter Tuning of Multimodal Optimization Performance by Means of Different Surrogate Models Catalin Stoean, Mike Preuss, Ruxandra Stoean

COFFEE BREAK

Session 4
16.10-16.35 ChallengA Behavior-based Analysis of Modal.
Leonardo Trujillo, Lee Spector, Enrique Naredo, Yuliana Martínez
16.40-17.05 Problem Understanding through Landscape Theory.
Francisco Chicano, Gabriel Luque, Enrique Alba
17.10-18.00 Discussion: Looking Ahead – An open discussion on the themes covered by the talks in the workshop as well as future development of the workshop series and research field

Details of the final programme will be posted shortly

Workshop Chairs

Emma Hart

Emma Hart

Prof Emma Hart Prof Emma Hart is an experienced researcher in the area of optmisation and hyper-heuristics, being one of the original authors of the term’hyper-heuristics’ in the early 2000s. She has published widely on the use of evolutionary algorithms applied to scheduling and timetabling problems, as well as in hyper-heuristics applied to a variety of domains. She is currently the holder of a large national research grant, investigating the use of hyper-heuristics that continuously learn and improve over time, which has collaborators from the logistics and forestry industry.

Kevin Sim

Kevin Sim

Kevin Sim is a researcher who works in the field of hyper-heuristics and problem-classification algorithms, and works with Prof. Hart on a large national grant on hyper-heuristics. He is also co-chair of a workshop at Evo-Star 2013 called EvoIndustry.

Organising Committee

Kent McClymont

Kent McClymont

Kent McClymont is an Associate Research Fellow in Computer Science at the University of Exeter. His research is focused on the study of optimisation problem analysis and multi-objective hyper-heuristic methods for solving hard real-world optimisation problems. Kent has a specific interest in heterogeneous encodings for which he has published a novel test problem suite and is continuing work in this area. He is a member of AISB committee and was is a member of the AISB 2013 Convention’s organising committee and has chaired sessions at previous GECCO and IEEE CEC.

Ed Keedwell

Ed Keedwell

Ed Keedwell is a Lecturer in Computer Science at the University of Exeter. His research is focused on Nature-Inspired Computation techniques and their application to real-world optimisation problems in engineering and bioinformatics. He has published almost 50 papers in this field and currently leads a group of 8-10 postgraduate students and postdoctoral researchers. Dr Keedwell is co-chair for the AISB2013 Convention and is organizer of a symposium for AISB 2013 and chaired sessions at previous AISB conventions and IEEE CEC 2010.

DOWNLOAD WORKSHOP PROGRAMME

Go to GECCO 2013 Website