A Formulation for Active Learning with Applications to Object Detection

We discuss a formulation for active example selection for function learning problems. This formulation is obtained by adapting Fedorov's optimal experiment design to the learning problem. We specifically show how to analytically derive example selection algorithms for certain well defined function classes. We then explore the behavior and sample complexity of such active learning algorithms. Finally, we view object detection as a special case of function learning and show how our formulation reduces to a useful heuristic to choose examples to reduce the generalization error.

Social Computing Data Repository

Social Computing Data Repository hosts data from a collection of many different social media sites, most of which have blogging capacity. Some of the prominent social media sites included in this repository are BlogCatalog, Twitter, MyBlogLog, Digg, StumbleUpon, del.icio.us, MySpace, LiveJournal, The Unofficial Apple Weblog (TUAW), Reddit, etc. The repository contains various facets of blog data including blog site metadata like, user defined tags, predefined categories, blog site description; blog post level metadata like, user defined tags, date and time of posting; blog posts; blog post mood (which is defined as the blogger's emotions when (s)he wrote the blog post); blogger name; blog post comments; and blogger social network.

Study of bond behaviour between FRP reinforcement and concrete

El uso de barras de materiales compuestos (FRP) se propone como una alternativa efectiva para las tradicionales estructuras de hormigón armadas con acero que sufren corrosión en ambientes agresivos. La aceptación de estos materiales en el mundo de la construcción está condicionada a la compresión de su comportamiento estructural. Este trabajo estudia el comportamiento adherente entre barras de FRP y hormigón mediante dos programas experimentales. El primero incluye la caracterización de la adherencia entre barras de FRP y hormigón mediante ensayos de pull-out y el segundo estudia el proceso de fisuración de tirantes de hormigón reforzados con barras de GFRP mediante ensayo a tracción directa. El trabajo se concluye con el desarrollo de un modelo numérico para la simulación del comportamiento de elementos de hormigón reforzado bajo cargas de tracción. La flexibilidad del modelo lo convierte en una herramienta flexible para la realización de un estudio paramétrico sobre las variables que influyen en el proceso de fisuración.

Algorithm for generating defective graphene sheets

An algorithm is presented for the generation of molecular models of defective graphene fragments, containing a majority of 6-membered rings with a small number of 5- and 7-membered rings as defects. The structures are generated from an initial random array of points in 2D space, which are then subject to Delaunay triangulation. The dual of the triangulation forms a Voronoi tessellation of polygons with a range of ring sizes. An iterative cycle of refinement, involving deletion and addition of points followed by further triangulation, is performed until the user-defined criteria for the number of defects are met. The array of points and connectivities are then converted to a molecular structure and subject to geometry optimization using a standard molecular modeling package to generate final atomic coordinates. On the basis of molecular mechanics with minimization, this automated method can generate structures, which conform to user-supplied criteria and avoid the potential bias associated with the manual building of structures. One application of the algorithm is the generation of structures for the evaluation of the reactivity of different defect sites. Ab initio electronic structure calculations on a representative structure indicate preferential fluorination close to 5-ring defects.

Application of complex extreme learning machine to multiclass classification problems with high dimensionality: a THz spectra classification problem

We extend extreme learning machine (ELM) classifiers to complex Reproducing Kernel Hilbert Spaces (RKHS) where the input/output variables as well as the optimization variables are complex-valued. A new family of classifiers, called complex-valued ELM (CELM) suitable for complex-valued multiple-input–multiple-output processing is introduced. In the proposed method, the associated Lagrangian is computed using induced RKHS kernels, adopting a Wirtinger calculus approach formulated as a constrained optimization problem similarly to the conventional ELM classifier formulation. When training the CELM, the Karush–Khun–Tuker (KKT) theorem is used to solve the dual optimization problem that consists of satisfying simultaneously smallest training error as well as smallest norm of output weights criteria. The proposed formulation also addresses aspects of quaternary classification within a Clifford algebra context. For 2D complex-valued inputs, user-defined complex-coupled hyper-planes divide the classifier input space into four partitions. For 3D complex-valued inputs, the formulation generates three pairs of complex-coupled hyper-planes through orthogonal projections. The six hyper-planes then divide the 3D space into eight partitions. It is shown that the CELM problem formulation is equivalent to solving six real-valued ELM tasks, which are induced by projecting the chosen complex kernel across the different user-defined coordinate planes. A classification example of powdered samples on the basis of their terahertz spectral signatures is used to demonstrate the advantages of the CELM classifiers compared to their SVM counterparts. The proposed classifiers retain the advantages of their ELM counterparts, in that they can perform multiclass classification with lower computational complexity than SVM classifiers. Furthermore, because of their ability to perform classification tasks fast, the proposed formulations are of interest to real-time applications.

A discontinuous-Galerkin-based immersed boundary method

A numerical method to approximate partial differential equations on meshes that do not conform to the domain boundaries is introduced. The proposed method is conceptually simple and free of user-defined parameters. Starting with a conforming finite element mesh, the key ingredient is to switch those elements intersected by the Dirichlet boundary to a discontinuous-Galerkin approximation and impose the Dirichlet boundary conditions strongly. By virtue of relaxing the continuity constraint at those elements. boundary locking is avoided and optimal-order convergence is achieved. This is shown through numerical experiments in reaction-diffusion problems. Copyright (c) 2008 John Wiley & Sons, Ltd.

Fault Coverage-Driven Incremental Test Generation

In this paper, we consider a classical problem of complete test generation for deterministic finite-state machines (FSMs) in a more general setting. The first generalization is that the number of states in implementation FSMs can even be smaller than that of the specification FSM. Previous work deals only with the case when the implementation FSMs are allowed to have the same number of states as the specification FSM. This generalization provides more options to the test designer: when traditional methods trigger a test explosion for large specification machines, tests with a lower, but yet guaranteed, fault coverage can still be generated. The second generalization is that tests can be generated starting with a user-defined test suite, by incrementally extending it until the desired fault coverage is achieved. Solving the generalized test derivation problem, we formulate sufficient conditions for test suite completeness weaker than the existing ones and use them to elaborate an algorithm that can be used both for extending user-defined test suites to achieve the desired fault coverage and for test generation. We present the experimental results that indicate that the proposed algorithm allows obtaining a trade-off between the length and fault coverage of test suites.

Formalization of an architectural model for exception handling coordination based on CA action concepts

Architectures based on Coordinated Atomic action (CA action) concepts have been used to build concurrent fault-tolerant systems. This conceptual model combines concurrent exception handling with action nesting to provide a general mechanism for both enclosing interactions among system components and coordinating forward error recovery measures. This article presents an architectural model to guide the formal specification of concurrent fault-tolerant systems. This architecture provides built-in Communicating Sequential Processes (CSPs) and predefined channels to coordinate exception handling of the user-defined components. Hence some safety properties concerning action scoping and concurrent exception handling can be proved by using the FDR (Failure Divergence Refinement) verification tool. As a result, a formal and general architecture supporting software fault tolerance is ready to be used and proved as users define components with normal and exceptional behaviors. (C) 2010 Elsevier B.V. All rights reserved.

Simulating autonomous robot teams with Microsoft robotics studio

This paper presents an application of Microsoft Robotics Studio (MSRS) in which a team of six four wheel drive, ground based robots explore and map simulated terrain. The user has the ability to modify the terrain and assign destination objectives to the team while the simulation is running. The terrain is initially generated using a gray scale image, in which the intensity of each pixel in the image gives an altitude datum. The robots start with no knowledge of their surroundings, and map the terrain as they attempt to reach user-defined target objectives. The mapping process simulates the use of common sensory hardware to determine datum points, including provision for field of view, detection range, and measurement accuracy. If traversal of a mapped area is indicated by the users’ targeting commands, path planning heuristics developed for MSRS by the author in earlier work are used to determine an efficient series of waypoints to reach the objective. Mutability of terrain is also explored- the user is able to modify the terrain without stopping the simulation. This forces the robots to adapt to changing environmental conditions, and permits analysis of the robustness of mapping algorithms used when faced with a changing world.

Modelling and optimising operations of a plant using JIT philosophy

An Australian automotive component company plans to assemble and deliver seats to a car manufacturer on a "just-in-time" basis at its new plant. The research objective was to model seat assembly operations and apply Toyota goal chasing algorithm and user defined algorithm to balance workload among all the assembly workstations and areas.

A study on mechanical behavior of functionally-graded carbon nanotube-reinforced nanocomposites

In this study, we are focusing on the investigation of the effects of gradient patterns on mechanical behavior of functionally-graded carbon nanotube-reinforced nanocomposites and considering typical beams made of such nanocomposites. Both analytic and finite element-based numerical models were developed. Analytic model was developed based on the first-order shear deformation and Timoshenko beam theories meanwhile finite element models were developed using Abaqus in conjunction with user-defined subroutines for defining the continuously gradient material properties for different gradient patterns. Position-dependent elastic modulus equations for four continuously graded patterns were studied. A nongraded pattern was used for benchmarking with the same geometry and total carbon nanotube (CNT) contents. For validation and verification, the results on both deflection and stress of these nanocomposite beams were analyzed, which clearly showed high influence from gradient patterns on these mechanical behaviors of such beams.

Parameterization Of Turbulence Models Using 3DVAR Data Assimilation

In this research the 3DVAR data assimilation scheme is implemented in the numerical model DIVAST in order to optimize the performance of the numerical model by selecting an appropriate turbulence scheme and tuning its parameters. Two turbulence closure schemes: the Prandtl mixing length model and the two-equation k-ε model were incorporated into DIVAST and examined with respect to their universality of application, complexity of solutions, computational efficiency and numerical stability. A square harbour with one symmetrical entrance subject to tide-induced flows was selected to investigate the structure of turbulent flows. The experimental part of the research was conducted in a tidal basin. A significant advantage of such laboratory experiment is a fully controlled environment where domain setup and forcing are user-defined. The research shows that the Prandtl mixing length model and the two-equation k-ε model, with default parameterization predefined according to literature recommendations, overestimate eddy viscosity which in turn results in a significant underestimation of velocity magnitudes in the harbour. The data assimilation of the model-predicted velocity and laboratory observations significantly improves model predictions for both turbulence models by adjusting modelled flows in the harbour to match de-errored observations. 3DVAR allows also to identify and quantify shortcomings of the numerical model. Such comprehensive analysis gives an optimal solution based on which numerical model parameters can be estimated. The process of turbulence model optimization by reparameterization and tuning towards optimal state led to new constants that may be potentially applied to complex turbulent flows, such as rapidly developing flows or recirculating flows.

An approach to implement data fusion techniques in wireless sensor networks using genetic machine learning algorithms

Wireless Sensor Networks (WSNs) can be used to monitor hazardous and inaccessible areas. In these situations, the power supply (e.g. battery) of each node cannot be easily replaced. One solution to deal with the limited capacity of current power supplies is to deploy a large number of sensor nodes, since the lifetime and dependability of the network will increase through cooperation among nodes. Applications on WSN may also have other concerns, such as meeting temporal deadlines on message transmissions and maximizing the quality of information. Data fusion is a well-known technique that can be useful for the enhancement of data quality and for the maximization of WSN lifetime. In this paper, we propose an approach that allows the implementation of parallel data fusion techniques in IEEE 802.15.4 networks. One of the main advantages of the proposed approach is that it enables a trade-off between different user-defined metrics through the use of a genetic machine learning algorithm. Simulations and field experiments performed in different communication scenarios highlight significant improvements when compared with, for instance, the Gur Game approach or the implementation of conventional periodic communication techniques over IEEE 802.15.4 networks. © 2013 Elsevier B.V. All rights reserved.

Estudo do método de inversão tomográfica 2D baseada nos atributos cinemáticos da onda NIP e construção de interface gráfica para o processo de picking

Neste trabalho, estuda-se um novo método de inversão tomográfica de reflexão para a determinação de um modelo isotrópico e suave de velocidade por meio da aplicação, em dados sintéticos e reais, do programa Niptomo que é uma implementação do método de inversão tomográfica dos atributos cinemáticos da onda hipotética do ponto de incidência normal (PIN). Os dados de entrada para a inversão tomográfica, isto é, o tempo de trânsito e os atributos da onda PIN (raio de curvatura da frente de onda emergente e ângulo de emergência), são retirados de uma série de pontos escolhidos na seção afastamento nulo (AN) simulada, obtida pelo método de empilhamento por superfícies de reflexão comum (SRC). Normalmente, a escolha destes pontos na seção AN é realizada utilizando-se programas de picking automático, que identificam eventos localmente coerentes na seção sísmica com base nos parâmetros fornecidos pelo usuário. O picking é um dos processos mais críticos dos métodos de inversão tomográfica, pois a inclusão de dados de eventos que não sejam de reflexões primárias podem ser incluídos neste processo, prejudicando assim o modelo de velocidades a ser obtido pela inversão tomográfica. Este trabalho tem por objetivo de construir um programa de picking interativo para fornecer ao usuário o controle da escolha dos pontos de reflexões sísmicas primárias, cujos dados serão utilizados na inversão tomográfica. Os processos de picking e inversão tomográfica são aplicados nos dados sintéticos Marmousi e nos dados da linha sísmica 50-RL-90 da Bacia do Tacutu. Os resultados obtidos mostraram que o picking interativo para a escolha de pontos sobre eventos de reflexões primárias favorece na obtenção de um modelo de velocidade mais preciso.

Desenvolvimento de uma metodologia de ensino, baseada na teoria das inteligências múltiplas viabilizada pelo uso de tecnologias de informação e comunicação

Pós-graduação em Engenharia Elétrica - FEIS