Botond Virginas

Expensive Multiobjective Optimization by MOEA/D With Gaussian Process Model

In some expensive multiobjective optimization problems (MOPs), several function evaluations can be carried out in a batch way. Therefore, it is very desirable to develop methods which can generate multipler test points simultaneously. This paper proposes such a method, called MOEA/D-EGO, for dealing with expensive multiobjective optimization. MOEA/D-EGO decomposes an MOP in question into a number of single-objective optimization subproblems. A predictive model is built for each subproblem based on the points evaluated so far. Effort has been made to reduce the overhead for modeling and to improve the prediction quality. At each generation, MOEA/D is used for maximizing the expected improvement metric values of all the subproblems, and then several test points are selected for evaluation. Extensive experimental studies have been carried out to investigate the ability of the proposed algorithm.

ARMS: An automated resource management system for British Telecommunications plc

Abstract Accurate demand forecasting combined with resource planning is critical to a company’s performance and profitability. This paper describes ARMS (automated resource management system), an integrated system developed for the customer service operations of British Telecommunications plc to help with the operational/tactical planning and deployment of the company’s 20,000-strong field engineer workforce. ARMS integrates a forecasting tool with a resource planning tool and a resource balancing tool providing an end-to-end automated resource management solution for the organisation. OR techniques are used throughout the system, including ARIMA for forecasting, constraint satisfaction for problem modelling, heuristic search for problem solving thus demonstrating the value and relevance of OR in solving today’s business problems.

Modular neural networks for recursive collaborative forecasting in the service chain

In order to honour customer demand and sustain quality of service in BTs service chain, accurate forecasting for customer demand is critical for optimal resource planning. In the more general context of service organisations, failure to allocate sufficient resources to meet anticipated customer demand will lead to delayed or disrupted service provision which in turn will result in degraded quality of service for customers and ill-balanced utilisation of available resources. In this paper, we present our ongoing research on a prototype collaborative forecasting application, whereas organisations involved in a supply and demand partnership aim to co-operate by sharing and jointly forming forecasts to aid in resource planning. We identify key theoretical and implementation specific issues related to the area of collaborative forecasting and discuss our initial modular artificial neural network approach to the problem.

On the performance of metamodel assisted MOEA/D

MOEA/D is a novel and successful Multi-Objective Evolutionary Algorithms(MOEA) which utilises the idea of problem decomposition to tackle the complexity from multiple objectives. It shows better performance than most nowadays mainstream MOEA methods in various test problems, especially on the quality of solutions distribution in the Pareto set. This paper aims to bring the strength of metamodel into MOEA/D to help the solving of expensive black-box multi-objective problems. Gaussian Random Field Metamodel(GRFM) is chosen as the approximation method. The performance is analysed and compared on several test problems, which shows a promising perspective on this method.

Retractable contract network for empowerment in workforce scheduling

This paper is about business modelling and negotiation protocol design in distributed scheduling, where individual agents have individual (potentially conflicting) interests. It is motivated by BTs workforce scheduling problem, in which multiple service providers have to serve multiple service buyers. The service providers and buyers all attempt to maximize their own utility. The overall problem is a multi-objective optimization problem; for example, one has to maximize completion rates and service quality and minimize travelling distances. Although the work is motivated by BTs business operations, the aim is to develop a general negotiation protocol for staff empowerment. Standard contract net is a practical strategy in distributed scheduling where agents may have conflicting objectives. In this paper, we have introduced a retractable contract net protocol, RECONNET, which supports hill-climbing in the space of schedules. It is built upon a job-release and compensation mechanism. A system based on RECONNET has been implemented for BTs workforce scheduling problem. The software, which we call ASMCR, allows the management to exert full control over the companys multiple objectives. The manager generates a Pareto set of solutions by defining, for each buyer and seller, the weights given to each objective. ASMCR gives service buyers and sellers ownership of their problem and freedom to maximize their performance under the criteria defined by the management. ASMCR was tested on real-sized problems and demonstrated to meet BTs operational time requirement. It has full potential to be further developed for tackling BTs workforce scheduling problem.

Fuzzy clustering based Gaussian Process Model for large training set and its application in expensive evolutionary optimization

Gaussian process model is an effective and efficient method for approximating a continuous function. However, its computational cost increases exponentially with the size of training data set. A very popular way to alleviate this shortcoming is to cluster the whole training data set into a number of small clusters and then a local model is built for each cluster. However, widely used crisp clustering might not be accurate in the boundary areas among different clusters. This paper proposes a fuzzy clustering based method for improving approximation quality. Several clusters with overlaps are firstly obtained by Fuzzy C-Means clustering and then local models are built for these clusters. It has been demonstrated that this method can be used with evolutionary algorithms for dealing expensive optimization problems.

Distributed Resource Allocation via Local Choices: General Model and a Basic Solution

This paper describes a solution to resource allocation, modelled as a distributed system. The solution was developed to complement an existing system built for the same purpose, but in a centralised approach (i.e., based on a central multi-criteria optimisation algorithm). Both are part of ARMS (Auto-mated Resource Management System), an integrated system for the customer service operations of British Telecommunications plc. Resource allocation is modelled here as an iterative communication process, based on a 4-step communication protocol, between resources agents, on one hand, and requirements agents, on the other. Agents “own” disjoint sets of resources or requirements – i.e., they have complete decision power over their allocation. The global allocation emerges from this interaction/communication, where choices/decisions are made locally, within each agent. The paper describes the main aspects of the distributed system and illustrates, with a concrete implementation, the emergence of a good global solution from the distributed algorithm.

Tchebycheff approximation in Gaussian Process model composition for multi-objective expensive black box

Black-box expensive function is ubiquitous in real world problems. Much research has been done on scalar objective optimization for such problems with great success. Comparatively, very little work has been done in multi-objective optimization. In many cases, it is not straightforward to convert methods from scalar objective optimization to multi-objective optimization due to the complexities incurred by Pareto domination. In our pervious research, concept of model composition based on Gaussian Process metamodel and the powerful MOEA/D framework proved to be a successful approach for multi-objective optimization with black-box expensive functions. We derived Weighted-Sum and Tchebycheff model composition for bi-objective problems. However, due to the complexity of Tchebycheff decomposition structure, it is very hard, if not impossible, to extend the method to three or more objective problems in a nature way. In this paper, we propose an approximation method for Tchebycheff model composition which greatly simplify the derivation for three or more objective cases. Experiments show the approximation produces very similar performance as the Weighted-Sum and Tchebycheff without approximation. Thus, the new method enables us to tackle multi-objective problems with black-box expensive functions that could not be tackled effectively so far.

A Two Stage Optimisation System for Resource Management in BT

Resource management is a three step process of job demand forecasting, resource planning and resource distribution. Job demand forecasting is data intensive and requires sophisticated algorithms. Resource planning and distribution are complex processes, usually involving the analyses of large amounts of information. The complexity increases when more than one objective is being evaluated and the number of variables to consider is huge. Clearly the amount of effort and time required for accurately managing resources warrant an automated approach, especially for large customer service organisations such as BT. This paper describes a two stage optimisation system which is part of the ARMS architecture for automating resource management in BT. The paper focuses on a hybrid resource planning and distribution system that is underpinned by constraint-based and multi-objective optimisation methods. We have developed Collaborator, a resource distribution system which sits on top of Dynamic Planner, a resource planning system. By using this novel integrated approach, BT aims to optimise resource deployment through increased workforce utilisation and mobility and ultimately to substantially reduce its operational costs.

Intelligent Resource Allocation---Solutions and Pathways in a Workforce Planning Problem

The paper is based on FieldExchange --- a computer system responsible for monitoring and supporting resource re-distribution decision making in BTs Operational Resource Management units. This paper considers the problem of resource allocation in the service industries approached from an agent-based perspective. The problem is formulated as a centralized/distributed planning problem. The paper describes the context of this solution, the general model and solution and four specific implementations with results and discussion.