Automatic blood glucose classification for gestational diabetes with feature selection: decision trees vs neural networks

Automatic blood glucose classification may help specialists to provide a better interpretation of blood glucose data, downloaded directly from patients glucose meter and will contribute in the development of decision support systems for gestational diabetes. This paper presents an automatic blood glucose classifier for gestational diabetes that compares 6 different feature selection methods for two machine learning algorithms: neural networks and decision trees. Three searching algorithms, Greedy, Best First and Genetic, were combined with two different evaluators, CSF and Wrapper, for the feature selection. The study has been made with 6080 blood glucose measurements from 25 patients. Decision trees with a feature set selected with the Wrapper evaluator and the Best first search algorithm obtained the best accuracy: 95.92%.

Global optimization of solar thermophotovoltaic systems

n this paper, we present a theoretical model based on the detailed balance theory of solar thermophotovoltaic systems comprising multijunction photovoltaic cells, a sunlight concentrator and spectrally selective surfaces. The full system has been defined by means of 2n + 8 variables (being n the number of sub-cells of the multijunction cell). These variables are as follows: the sunlight concentration factor, the absorber cut-off energy, the emitter-to-absorber area ratio, the emitter cut-off energy, the band-gap energy(ies) and voltage(s) of the sub-cells, the reflectivity of the cells' back-side reflector, the emitter-to-cell and cell-to-cell view factors and the emitter-to-cell area ratio. We have used this model for carrying out a multi-variable system optimization by means of a multidimensional direct-search algorithm. This analysis allows to find the set of system variables whose combined effects results in the maximum overall system efficiency. From this analysis, we have seen that multijunction cells are excellent candidates to enhance the system efficiency and the electrical power density. Particularly, multijunction cells report great benefits for systems with a notable presence of optical losses, which are unavoidable in practical systems. Also, we have seen that the use of spectrally selective absorbers, rather than black-body absorbers, allows to achieve higher system efficiencies for both lower concentration and lower emitter-to-absorber area ratio. Finally, we have seen that sun-to-electricity conversion efficiencies above 30% and electrical power densities above 50 W/cm2 are achievable for this kind of systems.

Avaliações e testes dos métodos MASW e ReMi por meio do tratamento de dados empíricos e sintéticos em um programa de processamento e inversão desenvolvido em MATLAB e sua implicação em um problema geológico na Bacia de Taubaté

Os métodos de ondas superficiais com ênfase nas ondas Rayleigh foram utilizados como o núcleo desse trabalho de Doutorado. Inicialmente, as ondas Rayleigh foram modeladas permitindo o estudo de sensibilidade de suas curvas de dispersão sob diferentes configurações de parâmetros físicos representando diversos modelos de camadas, em que pôde ser observado parâmetros com maior e menor sensibilidade e também alguns efeitos provocados por baixas razões de Poisson. Além disso, na fase de inversão dos dados a modelagem das ondas Rayleigh foi utilizada para a construção da função objeto, que agregada ao método de mínimos quadrados, a partir do método de Levenberg-Marquardt, permitiu a implementação de um algoritmo de busca local responsável pela inversão de dados das ondas superficiais. Por se tratar de um procedimento de busca local, o algoritmo de inversão foi complementado por uma etapa de pré-inversão com a geração de um modelo inicial para que o procedimento de inversão fosse mais rápido e eficiente. Visando uma eficiência ainda maior do procedimento de inversão, principalmente em modelos de camadas com inversão de velocidades, foi implementado um algoritmo de pós-inversão baseado em um procedimento de tentativa e erro minimizando os valores relativos da raiz quadrada do erro quadrático médio (REQMr) da inversão dos dados. Mais de 50 modelos de camadas foram utilizados para testar a modelagem, a pré-inversão, inversão e pós-inversão dos dados permitindo o ajuste preciso de parâmetros matemáticos e físicos presentes nos diversos scripts implementados em Matlab. Antes de inverter os dados adquiridos em campo, os mesmos precisaram ser tratados na etapa de processamento de dados, cujo objetivo principal é a extração da curva de dispersão originada devido às ondas superficiais. Para isso, foram implementadas, também em Matlab, três metodologias de processamento com abordagens matemáticas distintas. Essas metodologias foram testadas e avaliadas com dados sintéticos e reais em que foi possível constatar as virtudes e deficiências de cada metodologia estudada, bem como as limitações provocadas pela discretização dos dados de campo. Por último, as etapas de processamento, pré-inversão, inversão e pós-inversão dos dados foram unificadas para formar um programa de tratamento de dados de ondas superficiais (Rayleigh). Ele foi utilizado em dados reais originados pelo estudo de um problema geológico na Bacia de Taubaté em que foi possível mapear os contatos geológicos ao longo dos pontos de aquisição sísmica e compará-los a um modelo inicial existente baseado em observações geomorfológicas da área de estudos, mapa geológico da região e informações geológicas globais e locais dos movimentos tectônicos na região. As informações geofísicas associadas às geológicas permitiram a geração de um perfil analítico da região de estudos com duas interpretações geológicas confirmando a suspeita de neotectônica na região em que os contatos geológicos entre os depósitos Terciários e Quaternários foram identificados e se encaixaram no modelo inicial de hemi-graben com mergulho para Sudeste.

Toward efficient multifeature query processing

In many advanced applications, data are described by multiple high-dimensional features. Moreover, different queries may weight these features differently; some may not even specify all the features. In this paper, we propose our solution to support efficient query processing in these applications. We devise a novel representation that compactly captures f features into two components: The first component is a 2D vector that reflects a distance range ( minimum and maximum values) of the f features with respect to a reference point ( the center of the space) in a metric space and the second component is a bit signature, with two bits per dimension, obtained by analyzing each feature's descending energy histogram. This representation enables two levels of filtering: The first component prunes away points that do not share similar distance ranges, while the bit signature filters away points based on the dimensions of the relevant features. Moreover, the representation facilitates the use of a single index structure to further speed up processing. We employ the classical B+-tree for this purpose. We also propose a KNN search algorithm that exploits the access orders of critical dimensions of highly selective features and partial distances to prune the search space more effectively. Our extensive experiments on both real-life and synthetic data sets show that the proposed solution offers significant performance advantages over sequential scan and retrieval methods using single and multiple VA-files.

Purification of phosphoric acid by solvent extraction

This work follows a feasibility study (187) which suggested that a process for purifying wet-process phosphoric acid by solvent extraction should be economically viable. The work was divided into two main areas, (i) chemical and physical measurements on the three-phase system, with or without impurities; (ii) process simulation and optimization. The object was to test the process technically and economically and to optimise the type of solvent. The chemical equilibria and distribution curves for the system water - phosphoric acid - solvent for the solvents n-amyl alcohol, tri-n-butyl phosphate, di-isopropyl ether and methyl isobutyl ketone have been determined. Both pure phosphoric acid and acid containing known amounts of naturally occurring impurities (Fe P0 4 , A1P0 4 , Ca3(P04)Z and Mg 3(P0 4 )Z) were examined. The hydrodynamic characteristics of the systems were also studied. The experimental results obtained for drop size distribution were compared with those obtainable from Hinze's equation (32) and it was found that they deviated by an amount related to the turbulence. A comprehensive literature survey on the purification of wet-process phosphoric acid by organic solvents has been made. The literature regarding solvent extraction fundamentals and equipment and optimization methods for the envisaged process was also reviewed. A modified form of the Kremser-Brown and Souders equation to calculate the number of contact stages was derived. The modification takes into account the special nature of phosphoric acid distribution curves in the studied systems. The process flow-sheet was developed and simulated. Powell's direct search optimization method was selected in conjunction with the linear search algorithm of Davies, Swann and Campey. The objective function was defined as the total annual manufacturing cost and the program was employed to find the optimum operating conditions for anyone of the chosen solvents. The final results demonstrated the following order of feasibility to purify wet-process acid: di-isopropyl ether, methylisobutyl ketone, n-amyl alcohol and tri-n-butyl phosphate.

Vibration analysis of a printed circuit board

The reliability of the printed circuit board assembly under dynamic environments, such as those found onboard airplanes, ships and land vehicles is receiving more attention. This research analyses the dynamic characteristics of the printed circuit board (PCB) supported by edge retainers and plug-in connectors. By modelling the wedge retainer and connector as providing simply supported boundary condition with appropriate rotational spring stiffnesses along their respective edges with the aid of finite element codes, accurate natural frequencies for the board against experimental natural frequencies are obtained. For a PCB supported by two opposite wedge retainers and a plug-in connector and with its remaining edge free of any restraint, it is found that these real supports behave somewhere between the simply supported and clamped boundary conditions and provide a percentage fixity of 39.5% more than the classical simply supported case. By using an eigensensitivity method, the rotational stiffnesses representing the boundary supports of the PCB can be updated effectively and is capable of representing the dynamics of the PCB accurately. The result shows that the percentage error in the fundamental frequency of the PCB finite element model is substantially reduced from 22.3% to 1.3%. The procedure demonstrated the effectiveness of using only the vibration test frequencies as reference data when the mode shapes of the original untuned model are almost identical to the referenced modes/experimental data. When using only modal frequencies in model improvement, the analysis is very much simplified. Furthermore, the time taken to obtain the experimental data will be substantially reduced as the experimental mode shapes are not required.In addition, this thesis advocates a relatively simple method in determining the support locations for maximising the fundamental frequency of vibrating structures. The technique is simple and does not require any optimisation or sequential search algorithm in the analysis. The key to the procedure is to position the necessary supports at positions so as to eliminate the lower modes from the original configuration. This is accomplished by introducing point supports along the nodal lines of the highest possible mode from the original configuration, so that all the other lower modes are eliminated by the introduction of the new or extra supports to the structure. It also proposes inspecting the average driving point residues along the nodal lines of vibrating plates to find the optimal locations of the supports. Numerical examples are provided to demonstrate its validity. By applying to the PCB supported on its three sides by two wedge retainers and a connector, it is found that a single point constraint that would yield maximum fundamental frequency is located at the mid-point of the nodal line, namely, node 39. This point support has the effect of increasing the structure's fundamental frequency from 68.4 Hz to 146.9 Hz, or 115% higher.

Liquidity and portfolio optimisation

This thesis presents research within empirical financial economics with focus on liquidity and portfolio optimisation in the stock market. The discussion on liquidity is focused on measurement issues, including TAQ data processing and measurement of systematic liquidity factors (FSO). Furthermore, a framework for treatment of the two topics in combination is provided. The liquidity part of the thesis gives a conceptual background to liquidity and discusses several different approaches to liquidity measurement. It contributes to liquidity measurement by providing detailed guidelines on the data processing needed for applying TAQ data to liquidity research. The main focus, however, is the derivation of systematic liquidity factors. The principal component approach to systematic liquidity measurement is refined by the introduction of moving and expanding estimation windows, allowing for time-varying liquidity co-variances between stocks. Under several liability specifications, this improves the ability to explain stock liquidity and returns, as compared to static window PCA and market average approximations of systematic liquidity. The highest ability to explain stock returns is obtained when using inventory cost as a liquidity measure and a moving window PCA as the systematic liquidity derivation technique. Systematic factors of this setting also have a strong ability in explaining a cross-sectional liquidity variation. Portfolio optimisation in the FSO framework is tested in two empirical studies. These contribute to the assessment of FSO by expanding the applicability to stock indexes and individual stocks, by considering a wide selection of utility function specifications, and by showing explicitly how the full-scale optimum can be identified using either grid search or the heuristic search algorithm of differential evolution. The studies show that relative to mean-variance portfolios, FSO performs well in these settings and that the computational expense can be mitigated dramatically by application of differential evolution.

Transparent heterogeneous terrestrial optical communication networks with phase modulated signals

This thesis presents a large scale numerical investigation of heterogeneous terrestrial optical communications systems and the upgrade of fourth generation terrestrial core to metro legacy interconnects to fifth generation transmission system technologies. Retrofitting (without changing infrastructure) is considered for commercial applications. ROADM are crucial enabling components for future core network developments however their re-routing ability means signals can be switched mid-link onto sub-optimally configured paths which raises new challenges in network management. System performance is determined by a trade-off between nonlinear impairments and noise, where the nonlinear signal distortions depend critically on deployed dispersion maps. This thesis presents a comprehensive numerical investigation into the implementation of phase modulated signals in transparent reconfigurable wavelength division multiplexed fibre optic communication terrestrial heterogeneous networks. A key issue during system upgrades is whether differential phase encoded modulation formats are compatible with the cost optimised dispersion schemes employed in current 10 Gb/s systems. We explore how robust transmission is to inevitable variations in the dispersion mapping and how large the margins are when suboptimal dispersion management is applied. We show that a DPSK transmission system is not drastically affected by reconfiguration from periodic dispersion management to lumped dispersion mapping. A novel DPSK dispersion map optimisation methodology which reduces drastically the optimisation parameter space and the many ways to deploy dispersion maps is also presented. This alleviates strenuous computing requirements in optimisation calculations. This thesis provides a very efficient and robust way to identify high performing lumped dispersion compensating schemes for use in heterogeneous RZ-DPSK terrestrial meshed networks with ROADMs. A modified search algorithm which further reduces this number of configuration combinations is also presented. The results of an investigation of the feasibility of detouring signals locally in multi-path heterogeneous ring networks is also presented.

Entropy Based Approach to Finding Interacting Genes Responsible for Complex Human Disease

2000 Mathematics Subject Classification: 62P10, 92D10, 92D30, 94A17, 62L10.

Plaintext Recovery in DES-like Cryptosystems Based on S-boxes with Embedded Parity Check

We describe an approach for recovering the plaintext in block ciphers having a design structure similar to the Data Encryption Standard but with improperly constructed S-boxes. The experiments with a backtracking search algorithm performing this kind of attack against modified DES/Triple-DES in ECB mode show that the unknown plaintext can be recovered with a small amount of uncertainty and this algorithm is highly efficient both in time and memory costs for plaintext sources with relatively low entropy. Our investigations demonstrate once again that modifications resulting to S-boxes which still satisfy some design criteria may lead to very weak ciphers. ACM Computing Classification System (1998): E.3, I.2.7, I.2.8.

Parametric CAD model based shape optimization using adjoint functions

Adjoint methods have proven to be an efficient way of calculating the gradient of an objective function with respect to a shape parameter for optimisation, with a computational cost nearly independent of the number of the design variables [1]. The approach in this paper links the adjoint surface sensitivities (gradient of objective function with respect to the surface movement) with the parametric design velocities (movement of the surface due to a CAD parameter perturbation) in order to compute the gradient of the objective function with respect to CAD variables.
For a successful implementation of shape optimization strategies in practical industrial cases, the choice of design variables or parameterisation scheme used for the model to be optimized plays a vital role. Where the goal is to base the optimization on a CAD model the choices are to use a NURBS geometry generated from CAD modelling software, where the position of the NURBS control points are the optimisation variables [2] or to use the feature based CAD model with all of the construction history to preserve the design intent [3]. The main advantage of using the feature based model is that the optimized model produced can be directly used for the downstream applications including manufacturing and process planning.
This paper presents an approach for optimization based on the feature based CAD model, which uses CAD parameters defining the features in the model geometry as the design variables. In order to capture the CAD surface movement with respect to the change in design variable, the “Parametric Design Velocity” is calculated, which is defined as the movement of the CAD model boundary in the normal direction due to a change in the parameter value.
The approach presented here for calculating the design velocities represents an advancement in terms of capability and robustness of that described by Robinson et al. [3]. The process can be easily integrated to most industrial optimisation workflows and is immune to the topology and labelling issues highlighted by other CAD based optimisation processes. It considers every continuous (“real value”) parameter type as an optimisation variable, and it can be adapted to work with any CAD modelling software, as long as it has an API which provides access to the values of the parameters which control the model shape and allows the model geometry to be exported. To calculate the movement of the boundary the methodology employs finite differences on the shape of the 3D CAD models before and after the parameter perturbation. The implementation procedure includes calculating the geometrical movement along a normal direction between two discrete representations of the original and perturbed geometry respectively. Parametric design velocities can then be directly linked with adjoint surface sensitivities to extract the gradients to use in a gradient-based optimization algorithm.
The optimisation of a flow optimisation problem is presented, in which the power dissipation of the flow in an automotive air duct is to be reduced by changing the parameters of the CAD geometry created in CATIA V5. The flow sensitivities are computed with the continuous adjoint method for a laminar and turbulent flow [4] and are combined with the parametric design velocities to compute the cost function gradients. A line-search algorithm is then used to update the design variables and proceed further with optimisation process.

Guiding Monte Carlo tree searches with neural networks in the game of go

A dissertation submitted in fulfillment of the requirements to the degree of Master in Computer Science and Computer Engineering

Scheduling of track inspection and maintenance activities in railroad networks

The U.S. railroad companies spend billions of dollars every year on railroad track maintenance in order to ensure safety and operational efficiency of their railroad networks. Besides maintenance costs, other costs such as train accident costs, train and shipment delay costs and rolling stock maintenance costs are also closely related to track maintenance activities. Optimizing the track maintenance process on the extensive railroad networks is a very complex problem with major cost implications. Currently, the decision making process for track maintenance planning is largely manual and primarily relies on the knowledge and judgment of experts. There is considerable potential to improve the process by using operations research techniques to develop solutions to the optimization problems on track maintenance. In this dissertation study, we propose a range of mathematical models and solution algorithms for three network-level scheduling problems on track maintenance: track inspection scheduling problem (TISP), production team scheduling problem (PTSP) and job-to-project clustering problem (JTPCP). TISP involves a set of inspection teams which travel over the railroad network to identify track defects. It is a large-scale routing and scheduling problem where thousands of tasks are to be scheduled subject to many difficult side constraints such as periodicity constraints and discrete working time constraints. A vehicle routing problem formulation was proposed for TISP, and a customized heuristic algorithm was developed to solve the model. The algorithm iteratively applies a constructive heuristic and a local search algorithm in an incremental scheduling horizon framework. The proposed model and algorithm have been adopted by a Class I railroad in its decision making process. Real-world case studies show the proposed approach outperforms the manual approach in short-term scheduling and can be used to conduct long-term what-if analyses to yield managerial insights. PTSP schedules capital track maintenance projects, which are the largest track maintenance activities and account for the majority of railroad capital spending. A time-space network model was proposed to formulate PTSP. More than ten types of side constraints were considered in the model, including very complex constraints such as mutual exclusion constraints and consecution constraints. A multiple neighborhood search algorithm, including a decomposition and restriction search and a block-interchange search, was developed to solve the model. Various performance enhancement techniques, such as data reduction, augmented cost function and subproblem prioritization, were developed to improve the algorithm. The proposed approach has been adopted by a Class I railroad for two years. Our numerical results show the model solutions are able to satisfy all hard constraints and most soft constraints. Compared with the existing manual procedure, the proposed approach is able to bring significant cost savings and operational efficiency improvement. JTPCP is an intermediate problem between TISP and PTSP. It focuses on clustering thousands of capital track maintenance jobs (based on the defects identified in track inspection) into projects so that the projects can be scheduled in PTSP. A vehicle routing problem based model and a multiple-step heuristic algorithm were developed to solve this problem. Various side constraints such as mutual exclusion constraints and rounding constraints were considered. The proposed approach has been applied in practice and has shown good performance in both solution quality and efficiency.

Optimizing last mile delivery using public transport with multiagent based control

The majority of research work carried out in the field of Operations-Research uses methods and algorithms to optimize the pick-up and delivery problem. Most studies aim to solve the vehicle routing problem, to accommodate optimum delivery orders, vehicles etc. This paper focuses on green logistics approach, where existing Public Transport infrastructure capability of a city is used for the delivery of small and medium sized packaged goods thus, helping improve the situation of urban congestion and greenhouse gas emissions reduction. It carried out a study to investigate the feasibility of the proposed multi-agent based simulation model, for efficiency of cost, time and energy consumption. Multimodal Dijkstra Shortest Path algorithm and Nested Monte Carlo Search have been employed for a two-phase algorithmic approach used for generation of time based cost matrix. The quality of the tour is dependent on the efficiency of the search algorithm implemented for plan generation and route planning. The results reveal a definite advantage of using Public Transportation over existing delivery approaches in terms of energy efficiency.

Heterogeneous cooperative co-evolution memetic differential evolution algorithms for big data optimisation problems

Evolutionary algorithms (EAs) have recently been suggested as candidate for solving big data optimisation problems that involve very large number of variables and need to be analysed in a short period of time. However, EAs face scalability issue when dealing with big data problems. Moreover, the performance of EAs critically hinges on the utilised parameter values and operator types, thus it is impossible to design a single EA that can outperform all other on every problem instances. To address these challenges, we propose a heterogeneous framework that integrates a cooperative co-evolution method with various types of memetic algorithms. We use the cooperative co-evolution method to split the big problem into sub-problems in order to increase the efficiency of the solving process. The subproblems are then solved using various heterogeneous memetic algorithms. The proposed heterogeneous framework adaptively assigns, for each solution, different operators, parameter values and local search algorithm to efficiently explore and exploit the search space of the given problem instance. The performance of the proposed algorithm is assessed using the Big Data 2015 competition benchmark problems that contain data with and without noise. Experimental results demonstrate that the proposed algorithm, with the cooperative co-evolution method, performs better than without cooperative co-evolution method. Furthermore, it obtained very competitive results for all tested instances, if not better, when compared to other algorithms using a lower computational times.