'Explicit Learning: an Effort towards Human Scheduling Algorithms'

Abstract Scheduling problems are generally NP-hard combinatorial problems, and a lot of research has been done to solve these problems heuristically. However, most of the previous approaches are problem-specific and research into the development of a general scheduling algorithm is still in its infancy. Mimicking the natural evolutionary process of the survival of the fittest, Genetic Algorithms (GAs) have attracted much attention in solving difficult scheduling problems in recent years. Some obstacles exist when using GAs: there is no canonical mechanism to deal with constraints, which are commonly met in most real-world scheduling problems, and small changes to a solution are difficult. To overcome both difficulties, indirect approaches have been presented (in [1] and [2]) for nurse scheduling and driver scheduling, where GAs are used by mapping the solution space, and separate decoding routines then build solutions to the original problem. In our previous indirect GAs, learning is implicit and is restricted to the efficient adjustment of weights for a set of rules that are used to construct schedules. The major limitation of those approaches is that they learn in a non-human way: like most existing construction algorithms, once the best weight combination is found, the rules used in the construction process are fixed at each iteration. However, normally a long sequence of moves is needed to construct a schedule and using fixed rules at each move is thus unreasonable and not coherent with human learning processes. When a human scheduler is working, he normally builds a schedule step by step following a set of rules. After much practice, the scheduler gradually masters the knowledge of which solution parts go well with others. He can identify good parts and is aware of the solution quality even if the scheduling process is not completed yet, thus having the ability to finish a schedule by using flexible, rather than fixed, rules. In this research we intend to design more human-like scheduling algorithms, by using ideas derived from Bayesian Optimization Algorithms (BOA) and Learning Classifier Systems (LCS) to implement explicit learning from past solutions. BOA can be applied to learn to identify good partial solutions and to complete them by building a Bayesian network of the joint distribution of solutions [3]. A Bayesian network is a directed acyclic graph with each node corresponding to one variable, and each variable corresponding to individual rule by which a schedule will be constructed step by step. The conditional probabilities are computed according to an initial set of promising solutions. Subsequently, each new instance for each node is generated by using the corresponding conditional probabilities, until values for all nodes have been generated. Another set of rule strings will be generated in this way, some of which will replace previous strings based on fitness selection. If stopping conditions are not met, the Bayesian network is updated again using the current set of good rule strings. The algorithm thereby tries to explicitly identify and mix promising building blocks. It should be noted that for most scheduling problems the structure of the network model is known and all the variables are fully observed. In this case, the goal of learning is to find the rule values that maximize the likelihood of the training data. Thus learning can amount to 'counting' in the case of multinomial distributions. In the LCS approach, each rule has its strength showing its current usefulness in the system, and this strength is constantly assessed [4]. To implement sophisticated learning based on previous solutions, an improved LCS-based algorithm is designed, which consists of the following three steps. The initialization step is to assign each rule at each stage a constant initial strength. Then rules are selected by using the Roulette Wheel strategy. The next step is to reinforce the strengths of the rules used in the previous solution, keeping the strength of unused rules unchanged. The selection step is to select fitter rules for the next generation. It is envisaged that the LCS part of the algorithm will be used as a hill climber to the BOA algorithm. This is exciting and ambitious research, which might provide the stepping-stone for a new class of scheduling algorithms. Data sets from nurse scheduling and mall problems will be used as test-beds. It is envisaged that once the concept has been proven successful, it will be implemented into general scheduling algorithms. It is also hoped that this research will give some preliminary answers about how to include human-like learning into scheduling algorithms and may therefore be of interest to researchers and practitioners in areas of scheduling and evolutionary computation. References 1. Aickelin, U. and Dowsland, K. (2003) 'Indirect Genetic Algorithm for a Nurse Scheduling Problem', Computer & Operational Research (in print). 2. Li, J. and Kwan, R.S.K. (2003), 'Fuzzy Genetic Algorithm for Driver Scheduling', European Journal of Operational Research 147(2): 334-344. 3. Pelikan, M., Goldberg, D. and Cantu-Paz, E. (1999) 'BOA: The Bayesian Optimization Algorithm', IlliGAL Report No 99003, University of Illinois. 4. Wilson, S. (1994) 'ZCS: A Zeroth-level Classifier System', Evolutionary Computation 2(1), pp 1-18.

Exploração de escalonamentos óptimos na gestão de exames médicos

Uma das áreas de aplicação da optimização é a Engenharia Biomédica, pois a optimização intervém no estudo de próteses e implantes, na reconstrução tomográfica, na mecânica experimental, entre outras aplicações. Este projecto tem como principal objectivo a criação de um novo programa de marcação de exames médicos a fim de minimizar o tempo de espera na realização dos mesmos. É efectuada uma breve referência à teoria da optimização bem como à optimização linear e não-linear, aos algoritmos genéticos, que foram usados para a realização deste trabalho. É também apresentado um caso de estudo, formulado como um problema de optimização não linear com restrições. Com este estudo verificou-se que o escalonamento de exames médicos nunca poderá ser optimizado a 100por cento devido à quantidade de variáveis existentes, sendo que algumas delas não são passíveis de prever com antecedência.

Optimization of Enzymes Inactivation in High Pressure Processes

The demand of highest quality foods in terms of taste and their properties preservation without the use of additives is constantly increasing. Consequently, new approaches to food processing have been developed, as for example high-pressure technology which has proven to be very valuable because it allows to maintain good properties of food like some vitamins and, at the same time, to reduce some undesirable bacteria. This technology avoids the use of high temperatures during the process (not like Pasteurization), which may have adverse effect on some nutritional properties of the food, its flavour, etc. The models for some enzymatic inactivations, which depend on the pressure and temperature profiles are presented. This work deals with the optimization of the inactivation of certain enzymes when high pressure treatment on food processing is applied. The optimization algorithms will minimize the inactivation not only of a certain isolated enzyme but also to several enzymes that can be involved simultaneously in the high-pressure process.

Protein fold recognition without Boltzmann statistics or explicit physical basis

We present a fast method for finding optimal parameters for a low-resolution (threading) force field intended to distinguish correct from incorrect folds for a given protein sequence. In contrast to other methods, the parameterization uses information from >10(7) misfolded structures as well as a set of native sequence-structure pairs. In addition to testing the resulting force field's performance on the protein sequence threading problem, results are shown that characterize the number of parameters necessary for effective structure recognition.

Asymmetric MRI magnet design using a hybrid numerical method

This paper describes a hybrid numerical method for the design of asymmetric magnetic resonance imaging magnet systems. The problem is formulated as a field synthesis and the desired current density on the surface of a cylinder is first calculated by solving a Fredholm equation of the first kind. Nonlinear optimization methods are then invoked to fit practical magnet coils to the desired current density. The field calculations are performed using a semi-analytical method. A new type of asymmetric magnet is proposed in this work. The asymmetric MRI magnet allows the diameter spherical imaging volume to be positioned close to one end of the magnet. The main advantages of making the magnet asymmetric include the potential to reduce the perception of claustrophobia for the patient, better access to the patient by attending physicians, and the potential for reduced peripheral nerve stimulation due to the gradient coil configuration. The results highlight that the method can be used to obtain an asymmetric MRI magnet structure and a very homogeneous magnetic field over the central imaging volume in clinical systems of approximately 1.2 m in length. Unshielded designs are the focus of this work. This method is flexible and may be applied to magnets of other geometries. (C) 1999 Academic Press.

A hybrid, inverse approach to the design of magnetic resonance imaging magnets

This paper describes a hybrid numerical method of an inverse approach to the design of compact magnetic resonance imaging magnets. The problem is formulated as a field synthesis and the desired current density on the surface of a cylinder is first calculated by solving a Fredholm equation of the first, kind. Nonlinear optimization methods are then invoked to fit practical magnet coils to the desired current density. The field calculations are performed using a semi-analytical method. The emphasis of this work is on the optimal design of short MRI magnets. Details of the hybrid numerical model are presented, and the model is used to investigate compact, symmetric MRI magnets as well as asymmetric magnets. The results highlight that the method can be used to obtain a compact MRI magnet structure and a very homogeneous magnetic field over the central imaging volume in clinical systems of approximately 1 m in length, significantly shorter than current designs. Viable asymmetric magnet designs, in which the edge of the homogeneous region is very close to one end of the magnet system are also presented. Unshielded designs are the focus of this work. This method is flexible and may be applied to magnets of other geometries. (C) 2000 American Association of Physicists in Medicine. [S0094-2405(00)00303-5].

A design method for superconducting MRI magnets with ferromagnetic material

The emphasis of this work is on the optimal design of MRI magnets with both superconducting coils and ferromagnetic rings. The work is directed to the automated design of MRI magnet systems containing superconducting wire and both `cold' and `warm' iron. Details of the optimization procedure are given and the results show combined superconducting and iron material MRI magnets with excellent field characteristics. Strong, homogeneous central magnetic fields are produced with little stray or external field leakage. The field calculations are performed using a semi-analytical method for both current coil and iron material sources. Design examples for symmetric, open and asymmetric clinical MRI magnets containing both superconducting coils and ferromagnetic material are presented.

Optimização da lista de células vizinhas em redes LTE

Trabalho Final de Mestrado para obtenção do grau de Mestre em Engenharia de Electrónica e Telecomunicações

A multi-objective approach for the motion planning of redundant manipulators

Kinematic redundancy occurs when a manipulator possesses more degrees of freedom than those required to execute a given task. Several kinematic techniques for redundant manipulators control the gripper through the pseudo-inverse of the Jacobian, but lead to a kind of chaotic inner motion with unpredictable arm configurations. Such algorithms are not easy to adapt to optimization schemes and, moreover, often there are multiple optimization objectives that can conflict between them. Unlike single optimization, where one attempts to find the best solution, in multi-objective optimization there is no single solution that is optimum with respect to all indices. Therefore, trajectory planning of redundant robots remains an important area of research and more efficient optimization algorithms are needed. This paper presents a new technique to solve the inverse kinematics of redundant manipulators, using a multi-objective genetic algorithm. This scheme combines the closed-loop pseudo-inverse method with a multi-objective genetic algorithm to control the joint positions. Simulations for manipulators with three or four rotational joints, considering the optimization of two objectives in a workspace without and with obstacles are developed. The results reveal that it is possible to choose several solutions from the Pareto optimal front according to the importance of each individual objective.

Métodos de penalidade exacta para resolução de problemas de optimização não linear

In this work we present a classification of some of the existing Penalty Methods (denominated the Exact Penalty Methods) and describe some of its limitations and estimated. With these methods we can solve problems of optimization with continuous, discrete and mixing constrains, without requiring continuity, differentiability or convexity. The boarding consists of transforming the original problem, in a sequence of problems without constrains, derivate of the initial, making possible its resolution for the methods known for this type of problems. Thus, the Penalty Methods can be used as the first step for the resolution of constrained problems for methods typically used in by unconstrained problems. The work finishes discussing a new class of Penalty Methods, for nonlinear optimization, that adjust the penalty parameter dynamically.

A web-based tool to evaluate the iterative processes of direct search methods

Constrained and unconstrained Nonlinear Optimization Problems often appear in many engineering areas. In some of these cases it is not possible to use derivative based optimization methods because the objective function is not known or it is too complex or the objective function is non-smooth. In these cases derivative based methods cannot be used and Direct Search Methods might be the most suitable optimization methods. An Application Programming Interface (API) including some of these methods was implemented using Java Technology. This API can be accessed either by applications running in the same computer where it is installed or, it can be remotely accessed through a LAN or the Internet, using webservices. From the engineering point of view, the information needed from the API is the solution for the provided problem. On the other hand, from the optimization methods researchers’ point of view, not only the solution for the problem is needed. Also additional information about the iterative process is useful, such as: the number of iterations; the value of the solution at each iteration; the stopping criteria, etc. In this paper are presented the features added to the API to allow users to access to the iterative process data.

Direct-search penalty/barrier methods

In Nonlinear Optimization Penalty and Barrier Methods are normally used to solve Constrained Problems. There are several Penalty/Barrier Methods and they are used in several areas from Engineering to Economy, through Biology, Chemistry, Physics among others. In these areas it often appears Optimization Problems in which the involved functions (objective and constraints) are non-smooth and/or their derivatives are not know. In this work some Penalty/Barrier functions are tested and compared, using in the internal process, Derivative-free, namely Direct Search, methods. This work is a part of a bigger project involving the development of an Application Programming Interface, that implements several Optimization Methods, to be used in applications that need to solve constrained and/or unconstrained Nonlinear Optimization Problems. Besides the use of it in applied mathematics research it is also to be used in engineering software packages.

GAIC: Um sistema inteligente e flexível para simulação e apoio à participação de pequenos e médios consumidores na gestão ativa de cargas no âmbito de smart grids

A liberalização dos mercados de energia e a utilização intensiva de produção distribuída tem vindo a provocar uma alteração no paradigma de operação das redes de distribuição de energia elétrica. A continuidade da fiabilidade das redes de distribuição no contexto destes novos paradigmas requer alterações estruturais e funcionais. O conceito de Smart Grid vem permitir a adaptação das redes de distribuição ao novo contexto. Numa Smart Grid os pequenos e médios consumidores são chamados ao plano ativo das participações. Este processo é conseguido através da aplicação de programas de demand response e da existência de players agregadores. O uso de programas de demand response para alcançar benefícios para a rede encontra-se atualmente a ser estudado no meio científico. Porém, existe a necessidade de estudos que procurem benefícios para os pequenos e médios consumidores. O alcance dos benefícios para os pequenos e médios consumidores não é apenas vantajoso para o consumidor, como também o é para a rede elétrica de distribuição. A participação, dos pequenos e médios consumidores, em programas de demand response acontece significativamente através da redução de consumos energéticos. De modo a evitar os impactos negativos que podem provir dessas reduções, o trabalho aqui proposto faz uso de otimizações que recorrem a técnicas de aprendizagem através da utilização redes neuronais artificiais. Para poder efetuar um melhor enquadramento do trabalho com as Smart Grids, será desenvolvido um sistema multiagente capaz de simular os principais players de uma Smart Grid. O foco deste sistema multiagente será o agente responsável pela simulação do pequeno e médio consumidor. Este agente terá não só que replicar um pequeno e médio consumidor, como terá ainda que possibilitar a integração de cargas reais e virtuais. Como meio de interação com o pequeno e médio consumidor, foi desenvolvida no âmbito desta dissertação um sistema móvel. No final do trabalho obteve-se um sistema multiagente capaz de simular uma Smart Grid e a execução de programas de demand response, sSendo o agente representante do pequeno e médio consumidor capaz de tomar ações e reações de modo a poder responder autonomamente aos programas de demand response lançados na rede. O desenvolvimento do sistema permite: o estudo e análise da integração dos pequenos e médios consumidores nas Smart Grids por meio de programas de demand response; a comparação entre múltiplos algoritmos de otimização; e a integração de métodos de aprendizagem. De modo a demonstrar e viabilizar as capacidades de todo o sistema, a dissertação inclui casos de estudo para as várias vertentes que podem ser exploradas com o sistema desenvolvido.

Retroanálise de parâmetros geomecânicos baseada em algoritmos de otimização

Dissertação de mestrado integrado em Engenharia Civil