Driving licensing renewal policy using neural network-based probabilistic decision support system

This paper investigates neural network-based probabilistic decision support system to assess drivers' knowledge for the objective of developing a renewal policy of driving licences. The probabilistic model correlates drivers' demographic data to their results in a simulated written driving exam (SWDE). The probabilistic decision support system classifies drivers' into two groups of passing and failing a SWDE. Knowledge assessment of drivers within a probabilistic framework allows quantifying and incorporating uncertainty information into the decision-making system. The results obtained in a Jordanian case study indicate that the performance of the probabilistic decision support systems is more reliable than conventional deterministic decision support systems. Implications of the proposed probabilistic decision support systems on the renewing of the driving licences decision and the possibility of including extra assessment methods are discussed.

Modelling question selection behaviour in online communities

Value of online Question Answering (QandA) communities is driven by the question-answering behaviour of its members. Finding the questions that members are willing to answer is therefore vital to the effcient operation of such communities. In this paper, we aim to identify the parameters that cor- relate with such behaviours. We train different models and construct effective predictions using various user, question and thread feature sets. We show that answering behaviour can be predicted with a high level of success.

Using observation ageing to improve Markovian model learning in QoS engineering

Markovian models are widely used to analyse quality-of-service properties of both system designs and deployed systems. Thanks to the emergence of probabilistic model checkers, this analysis can be performed with high accuracy. However, its usefulness is heavily dependent on how well the model captures the actual behaviour of the analysed system. Our work addresses this problem for a class of Markovian models termed discrete-time Markov chains (DTMCs). We propose a new Bayesian technique for learning the state transition probabilities of DTMCs based on observations of the modelled system. Unlike existing approaches, our technique weighs observations based on their age, to account for the fact that older observations are less relevant than more recent ones. A case study from the area of bioinformatics workflows demonstrates the effectiveness of the technique in scenarios where the model parameters change over time.

Deep Head Pose: Gaze-Direction Estimation in Multimodal Video

In this paper we present a convolutional neuralnetwork (CNN)-based model for human head pose estimation inlow-resolution multi-modal RGB-D data. We pose the problemas one of classification of human gazing direction. We furtherfine-tune a regressor based on the learned deep classifier. Next wecombine the two models (classification and regression) to estimateapproximate regression confidence. We present state-of-the-artresults in datasets that span the range of high-resolution humanrobot interaction (close up faces plus depth information) data tochallenging low resolution outdoor surveillance data. We buildupon our robust head-pose estimation and further introduce anew visual attention model to recover interaction with theenvironment. Using this probabilistic model, we show thatmany higher level scene understanding like human-human/sceneinteraction detection can be achieved. Our solution runs inreal-time on commercial hardware

Human behaviour recognition in data-scarce domains

This paper presents the novel theory for performing multi-agent activity recognition without requiring large training corpora. The reduced need for data means that robust probabilistic recognition can be performed within domains where annotated datasets are traditionally unavailable. Complex human activities are composed from sequences of underlying primitive activities. We do not assume that the exact temporal ordering of primitives is necessary, so can represent complex activity using an unordered bag. Our three-tier architecture comprises low-level video tracking, event analysis and high-level inference. High-level inference is performed using a new, cascading extension of the Rao–Blackwellised Particle Filter. Simulated annealing is used to identify pairs of agents involved in multi-agent activity. We validate our framework using the benchmarked PETS 2006 video surveillance dataset and our own sequences, and achieve a mean recognition F-Score of 0.82. Our approach achieves a mean improvement of 17% over a Hidden Markov Model baseline.

MixKMeans: Clustering Question-Answer Archives

Community-driven Question Answering (CQA) systems that crowdsource experiential information in the form of questions and answers and have accumulated valuable reusable knowledge. Clustering of QA datasets from CQA systems provides a means of organizing the content to ease tasks such as manual curation and tagging. In this paper, we present a clustering method that exploits the two-part question-answer structure in QA datasets to improve clustering quality. Our method, {\it MixKMeans}, composes question and answer space similarities in a way that the space on which the match is higher is allowed to dominate. This construction is motivated by our observation that semantic similarity between question-answer data (QAs) could get localized in either space. We empirically evaluate our method on a variety of real-world labeled datasets. Our results indicate that our method significantly outperforms state-of-the-art clustering methods for the task of clustering question-answer archives.

Online MBPeT dashboard : enabling performance testing as a service in the cloud

In this thesis, tool support is addressed for the combined disciplines of Model-based testing and performance testing. Model-based testing (MBT) utilizes abstract behavioral models to automate test generation, thus decreasing time and cost of test creation. MBT is a functional testing technique, thereby focusing on output, behavior, and functionality. Performance testing, however, is non-functional and is concerned with responsiveness and stability under various load conditions. MBPeT (Model-Based Performance evaluation Tool) is one such tool which utilizes probabilistic models, representing dynamic real-world user behavior patterns, to generate synthetic workload against a System Under Test and in turn carry out performance analysis based on key performance indicators (KPI). Developed at Åbo Akademi University, the MBPeT tool is currently comprised of a downloadable command-line based tool as well as a graphical user interface. The goal of this thesis project is two-fold: 1) to extend the existing MBPeT tool by deploying it as a web-based application, thereby removing the requirement of local installation, and 2) to design a user interface for this web application which will add new user interaction paradigms to the existing feature set of the tool. All phases of the MBPeT process will be realized via this single web deployment location including probabilistic model creation, test configurations, test session execution against a SUT with real-time monitoring of user configurable metric, and final test report generation and display. This web application (MBPeT Dashboard) is implemented with the Java programming language on top of the Vaadin framework for rich internet application development. The Vaadin framework handles the complicated web communications processes and front-end technologies, freeing developers to implement the business logic as well as the user interface in pure Java. A number of experiments are run in a case study environment to validate the functionality of the newly developed Dashboard application as well as the scalability of the solution implemented in handling multiple concurrent users. The results support a successful solution with regards to the functional and performance criteria defined, while improvements and optimizations are suggested to increase both of these factors.

Probabilistic verification of satellite systems for mission critical applications

In this thesis, we present a quantitative approach using probabilistic verification techniques for the analysis of reliability, availability, maintainability, and safety (RAMS) properties of satellite systems. The subject of our research is satellites used in mission critical industrial applications. A strong case for using probabilistic model checking to support RAMS analysis of satellite systems is made by our verification results. This study is intended to build a foundation to help reliability engineers with a basic background in model checking to apply probabilistic model checking to small satellite systems. We make two major contributions. One of these is the approach of RAMS analysis to satellite systems. In the past, RAMS analysis has been extensively applied to the field of electrical and electronics engineering. It allows system designers and reliability engineers to predict the likelihood of failures from the indication of historical or current operational data. There is a high potential for the application of RAMS analysis in the field of space science and engineering. However, there is a lack of standardisation and suitable procedures for the correct study of RAMS characteristics for satellite systems. This thesis considers the promising application of RAMS analysis to the case of satellite design, use, and maintenance, focusing on its system segments. Data collection and verification procedures are discussed, and a number of considerations are also presented on how to predict the probability of failure. Our second contribution is leveraging the power of probabilistic model checking to analyse satellite systems. We present techniques for analysing satellite systems that differ from the more common quantitative approaches based on traditional simulation and testing. These techniques have not been applied in this context before. We present the use of probabilistic techniques via a suite of detailed examples, together with their analysis. Our presentation is done in an incremental manner: in terms of complexity of application domains and system models, and a detailed PRISM model of each scenario. We also provide results from practical work together with a discussion about future improvements.

SEQUENTIAL DECISIONS AND PREDICTIONS IN NATURAL LANGUAGE PROCESSING

Natural language processing has achieved great success in a wide range of ap- plications, producing both commercial language services and open-source language tools. However, most methods take a static or batch approach, assuming that the model has all information it needs and makes a one-time prediction. In this disser- tation, we study dynamic problems where the input comes in a sequence instead of all at once, and the output must be produced while the input is arriving. In these problems, predictions are often made based only on partial information. We see this dynamic setting in many real-time, interactive applications. These problems usually involve a trade-off between the amount of input received (cost) and the quality of the output prediction (accuracy). Therefore, the evaluation considers both objectives (e.g., plotting a Pareto curve). Our goal is to develop a formal understanding of sequential prediction and decision-making problems in natural language processing and to propose efficient solutions. Toward this end, we present meta-algorithms that take an existent batch model and produce a dynamic model to handle sequential inputs and outputs. Webuild our framework upon theories of Markov Decision Process (MDP), which allows learning to trade off competing objectives in a principled way. The main machine learning techniques we use are from imitation learning and reinforcement learning, and we advance current techniques to tackle problems arising in our settings. We evaluate our algorithm on a variety of applications, including dependency parsing, machine translation, and question answering. We show that our approach achieves a better cost-accuracy trade-off than the batch approach and heuristic-based decision- making approaches. We first propose a general framework for cost-sensitive prediction, where dif- ferent parts of the input come at different costs. We formulate a decision-making process that selects pieces of the input sequentially, and the selection is adaptive to each instance. Our approach is evaluated on both standard classification tasks and a structured prediction task (dependency parsing). We show that it achieves similar prediction quality to methods that use all input, while inducing a much smaller cost. Next, we extend the framework to problems where the input is revealed incremen- tally in a fixed order. We study two applications: simultaneous machine translation and quiz bowl (incremental text classification). We discuss challenges in this set- ting and show that adding domain knowledge eases the decision-making problem. A central theme throughout the chapters is an MDP formulation of a challenging problem with sequential input/output and trade-off decisions, accompanied by a learning algorithm that solves the MDP.

Context Models for Understanding Images and Videos

A computer vision system that has to interact in natural language needs to understand the visual appearance of interactions between objects along with the appearance of objects themselves. Relationships between objects are frequently mentioned in queries of tasks like semantic image retrieval, image captioning, visual question answering and natural language object detection. Hence, it is essential to model context between objects for solving these tasks. In the first part of this thesis, we present a technique for detecting an object mentioned in a natural language query. Specifically, we work with referring expressions which are sentences that identify a particular object instance in an image. In many referring expressions, an object is described in relation to another object using prepositions, comparative adjectives, action verbs etc. Our proposed technique can identify both the referred object and the context object mentioned in such expressions. Context is also useful for incrementally understanding scenes and videos. In the second part of this thesis, we propose techniques for searching for objects in an image and events in a video. Our proposed incremental algorithms use the context from previously explored regions to prioritize the regions to explore next. The advantage of incremental understanding is restricting the amount of computation time and/or resources spent for various detection tasks. Our first proposed technique shows how to learn context in indoor scenes in an implicit manner and use it for searching for objects. The second technique shows how explicitly written context rules of one-on-one basketball can be used to sequentially detect events in a game.

¿Costumbre o Reacción? El efecto de ser víctima de delitos y corrupción en la satisfacción con la vida de los colombianos

Colombia ha sido un país con una larga historia de violencia y corrupción, por lo que se hace importante analizar la relación entre la victimización de dichos delitos sufrida por los individuos y su percepción de satisfacción con la vida. Se utiliza información entre 2004 y 2014 contenida en LAPOP, a manera de secciones transversales repetidas. Con el fin de encontrar dicho efecto, se estima robustamente un modelo probabilístico ordenado, en donde los resultados sugieren que la victimización del último año reduce en 6.7 puntos porcentuales la probabilidad de sentirse muy satisfecho con la vida y el haber sido víctima de algún soborno en 5 puntos porcentuales.

Diagnostic errors and repetitive sequential classifications in on-line process control by attributes

The procedure of on-line process control by attributes, known as Taguchi`s on-line process control, consists of inspecting the mth item (a single item) at every m produced items and deciding, at each inspection, whether the fraction of conforming items was reduced or not. If the inspected item is nonconforming, the production is stopped for adjustment. As the inspection system can be subject to diagnosis errors, one develops a probabilistic model that classifies repeatedly the examined item until a conforming or b non-conforming classification is observed. The first event that occurs (a conforming classifications or b non-conforming classifications) determines the final classification of the examined item. Proprieties of an ergodic Markov chain were used to get the expression of average cost of the system of control, which can be optimized by three parameters: the sampling interval of the inspections (m); the number of repeated conforming classifications (a); and the number of repeated non-conforming classifications (b). The optimum design is compared with two alternative approaches: the first one consists of a simple preventive policy. The production system is adjusted at every n produced items (no inspection is performed). The second classifies the examined item repeatedly r (fixed) times and considers it conforming if most classification results are conforming. Results indicate that the current proposal performs better than the procedure that fixes the number of repeated classifications and classifies the examined item as conforming if most classifications were conforming. On the other hand, the preventive policy can be averagely the most economical alternative rather than those ones that require inspection depending on the degree of errors and costs. A numerical example illustrates the proposed procedure. (C) 2009 Elsevier B. V. All rights reserved.

Economic design of an X chart for short-run production

The aim of this paper is to present an economical design of an X chart for a short-run production. The process mean starts equal to mu(0) (in-control, State I) and in a random time it shifts to mu(1) > mu(0) (out-of-control, State II). The monitoring procedure consists of inspecting a single item at every m produced ones. If the measurement of the quality characteristic does not meet the control limits, the process is stopped, adjusted, and additional (r - 1) items are inspected retrospectively. The probabilistic model was developed considering only shifts in the process mean. A direct search technique is applied to find the optimum parameters which minimizes the expected cost function. Numerical examples illustrate the proposed procedure. (C) 2009 Elsevier B.V. All rights reserved.

Numerical simulation of dendrite growth during solidification

A comprehensive probabilistic model for simulating dendrite morphology and investigating dendritic growth kinetics during solidification has been developed, based on a modified Cellular Automaton (mCA) for microscopic modeling of nucleation, growth of crystals and solute diffusion. The mCA model numerically calculated solute redistribution both in the solid and liquid phases, the curvature of dendrite tips and the growth anisotropy. This modeling takes account of thermal, curvature and solute diffusion effects. Therefore, it can simulate microstructure formation both on the scale of the dendrite tip length. This model was then applied for simulating dendritic solidification of an Al-7%Si alloy. Both directional and equiaxed dendritic growth has been performed to investigate the growth anisotropy and cooling rate on dendrite morphology. Furthermore, the competitive growth and selection of dendritic crystals have also investigated.

A practical approach to assess depression risk and to guide risk reduction strategies in later life

Background: Many factors have been associated with the onset and maintenance of depressive symptoms in later life, although this knowledge is yet to be translated into significant health gains for the population. This study gathered information about common modifiable and non-modifiable risk factors for depression with the aim of developing a practical probabilistic model of depression that can be used to guide risk reduction strategies. \Methods: A cross-sectional study was undertaken of 20,677 community-dwelling Australians aged 60 years or over in contact with their general practitioner during the preceding 12 months. Prevalent depression (minor or major) according to the Patient Health Questionnaire (PHQ-9) assessment was the main outcome of interest. Other measured exposures included self-reported age, gender, education, loss of mother or father before age 15 years, physical or sexual abuse before age 15 years, marital status, financial stress, social support, smoking and alcohol use, physical activity, obesity, diabetes, hypertension, and prevalent cardiovascular diseases, chronic respiratory diseases and cancer. Results: The mean age of participants was 71.7 +/- 7.6 years and 57.9% were women. Depression was present in 1665 (8.0%) of our subjects. Multivariate logistic regression showed depression was independently associated with age older than 75 years, childhood adverse experiences, adverse lifestyle practices (smoking, risk alcohol use, physical inactivity), intermediate health hazards (obesity, diabetes and hypertension), comorbid medical conditions (clinical history of coronary heart disease, stroke, asthma, chronic obstructive pulmonary disease, emphysema or cancers), and social or financial strain. We stratified the exposures to build a matrix that showed that the probability of depression increased progressively with the accumulation of risk factors, from less than 3% for those with no adverse factors to more than 80% for people reporting the maximum number of risk factors. Conclusions: Our probabilistic matrix can be used to estimate depression risk and to guide the introduction of risk reduction strategies. Future studies should now aim to clarify whether interventions designed to mitigate the impact of risk factors can change the prevalence and incidence of depression in later life.