Sign Out
Logged In:
Tab Image

Our Conferences / Events

Beale Lecture
The Royal Society, London
21 February 2019

New to OR 2019 Biennial Conference
Conference Aston Meeting Suites (CAMS) at Aston University Business School
10 - 11 April 2019

2nd IMA & OR Society Conference on Mathematics of Operational Research
Aston University, Birmingham
Thursday 25 - Friday 26 April 2019

OR61: Annual Conference
University of Kent
3 - 5 September 2019

Tab Image

Previous Beale Lectures - Beale Lecture 2017


This open event took place on Thursday 2nd March 2017 and began with a short talk from our 2014 PhD winner, Martin Takac, followed by a talk from our main speaker and 2015 Beale Medal Award winner, John Friend.

Title:  Connecting Public Policy Worlds.

Beale Medal Winner 2015 – Mr John Friend

The intricate webs of accountability involved in shaping public policy choices present structural challenges to policy makers and their advisers; and these in turn create opportunities to introduce innovative approaches to the design and facilitation of collaborative policy processes.  Numerous instances have now been reported, from Europe and further afield, of successful engagement between management scientists and public servants in designing and facilitating interactive processes for developing public policies in national, regional and local forums. 

One of the most significant influences has been the legacy of innovation that began with the launch over 50 years ago of a strategic partnership between our national OR Society and the Tavistock Institute of Human Relations.  John Friend will describe how the pioneering work of the centre which was known as the Institute for OR [IOR] has led to a widely-applied toolkit of interactive methods for structuring policy choices and sustaining strategic progress in the face of competing sources of uncertainty and urgency.  A further outcome has been a set of foundations for a useful science of public policy choice in multi-organisational domains; these have attracted international interest within academic and professional communities which have few historic links to OR.

The impetus has now gradually passed to consultants and public servants engaged in the practice of developing public policies in several different parts of the world.  This leads to the important question of whether younger generations of management scientists will be able to maintain the unifying drive and the momentum of innovation that have been so essential to the progress achieved so far. 

It will be argued that UK universities are now strategically placed to develop links with international partners in promoting the further development of a useful science of public policy design.  This however will depend on a readiness within one or more ambitious academic schools of management and decision science to forge collaborative links with sister schools that engage with communities of practice in important and interdependent public policy worlds.

John Friend joined the OR Society in 1957, and was elected a Companion of OR in 2008.  After graduating in mathematics and spending ten years in analytical roles in industry and air transport, he encountered quite different challenges on joining the Institute for OR, set up in 1963 as a joint venture of the OR Society and the Tavistock Institute to expand the horizons of OR through collaboration with social scientists in addressing important issues of public policy.He is best known in the OR world for his pioneering role in developing the Strategic Choice Approach, now viewed as a leading member of the softer OR family of problem structuring methods.  He has also attracted international attention among policy professionals and academics as a source of fresh insights into the inter-organisational dynamics of public policy choice.  Meanwhile, in the world of environmental planning he has become recognised for his contributions, alongside associates in many parts of the world, to the introduction of flexible processes of interactive planning under uncertainty at levels from the local community to the nation state. 

John Friend is the lead author of three books on these topics and the designer of interactive software for supporting strategic decision processes.  He has held visiting professorships at four British universities and was awarded an honorary doctorate by the University of Amsterdam.

Title: Randomized Coordinate Descent Methods for Big Data Optimizaton

PhD Winner 2014 – Dr Martin Takac

For ‘The Most Distinguished Body of Research leading to the Award of a Doctorate in the field of O.R.’ 

In today’s digital world there is an ever increasing demand for solving “big data” problems, each described by gigabytes or terabytes of data from sources such as twitter feeds, online image databases, text corpora, videos, government records, scientific experiments or click behaviour of online users. In many applications the problem at hand is formulated as an optimization problem: we seek to determine a number of variables minimizing a certain loss function (or maximizing a profit function), subject to constraints.

Classical optimization algorithms tend to require only a small number of iterations and output high precision result, but this comes at the cost of data-heavy iterations. For example,  interior point methods require the solution of an n × n system of equations, which in  general takes O(n3) floating point operations (additions or multiplications). If n = 109 for instance, Gaussian elimination  would require 471 years of computation on the currently ranked #1 supercomputer in the world (Tianhe-2) using all of its 3,120,000 cores. In the big data setting it is often prohibitive or downright impossible to perform even a single iteration of the classical methods, and hence focus is shifting to algorithms which are able to use less information to produce an iteration, thus making the iterative process feasible. This comes at the cost of increasing the number of iterations. In Truss Topology Design, for instance, each iteration of a coordinate descent method (CDM) consists of as little as 8 multiplications and 8 additions, and does not depend on the size of the problem. On the other hand, the number of iterations will depend on the problem size.

In this talk we analyse iteration complexity (how many iterations of the algorithm is sufficient to obtain sufficient solution) of coordinate descent methods for various loss functions. Moreover, in order to make use of the modern high-performance computers, the parallel version of CDM is proposed and analysed.

Martin Takác was born in Slovakia in 1986. He received the B.Sc. and M.Sc. degree in Economic and Financial Mathematics from Comenius University in Bratislava, and  Ph.D. degrees from University of Edinburgh, Scotland, United Kingdom, in 2008, 2010, and 2014, respectively. In 2014, he joined the Department of Industrial and Systems Engineering, Lehigh University, as an Assistant Professor. His current research interests include convex and non-convex optimization, design and analysis of algorithms, parallel and distributed computing, GPU computing and machine learning.