An Intelligent Driver Warning System

This chapter is focussed on the research and development of an intelligent driver warning system (IDWS) as a means to improve road safety and driving comfort. Two independent IDWS case studies are presented. The first study examines the methodology and implementation for attentive visual tracking and trajectory estimation for dynamic scene segmentation problems. In the second case study, the concept of driver modelling is evaluated which can be used to provide useful feedback to drivers. In both case studies, the quality of IDWS is largely determined by the modelling capability for estimating multiple vehicle trajectories and modelling driving behaviour. A class of modelling techniques based on neural-fuzzy systems, which exhibits provable learning and modelling capability, is proposed. For complex modelling problems where the curse of dimensionality becomes an issue, a network construction algorithm based on Adaptive Spline Modelling of Observation Data (ASMOD) is also proposed.

Semi-supervised document clustering via loci

Document clustering is one of the prominent methods for mining important information from the vast amount of data available on the web. However, document clustering generally suffers from the curse of dimensionality. Providentially in high dimensional space, data points tend to be more concentrated in some areas of clusters. We take advantage of this phenomenon by introducing a novel concept of dynamic cluster representation named as loci. Clusters’ loci are efficiently calculated using documents’ ranking scores generated from a search engine. We propose a fast loci-based semi-supervised document clustering algorithm that uses clusters’ loci instead of conventional centroids for assigning documents to clusters. Empirical analysis on real-world datasets shows that the proposed method produces cluster solutions with promising quality and is substantially faster than several benchmarked centroid-based semi-supervised document clustering methods.

Evaluation of a workplace alcohol and other drug policy

BACKGROUND The current impetus for developing alcohol and/or other drugs (AODs) workplace policies in Australia is to reduce workplace AOD impairment, improve safety, and prevent AOD-related injury in the workplace. For these policies to be effective, they need to be informed by scientific evidence. Evidence to inform the development and implementation of effective workplace AOD policies is currently lacking. There does not currently appear to be conclusive evidence for the effectiveness of workplace AOD policies in reducing impairment and preventing AOD-related injury. There is also no apparent evidence regarding which factors facilitate or impede the success of an AOD policy, or whether, for example, unsuccessful policy outcomes were due to poor policy or merely poor implementation of the policy. It was the aim of this research to undertake a process, impact, and outcome evaluation of a workplace AOD policy, and to contribute to the body of knowledge on the development and implementation of effective workplace AOD policies. METHODS The research setting was a state-based power-generating industry in Australia between May 2008 and May 2010. Participants for the process evaluation study were individuals who were integral to either the development or the implementation of the workplace AOD policy, or both of these processes (key informants), and comprised the majority of individuals who were involved in the process of developing and/or implementing the workplace AOD policy. The sample represented the two main groups of interest—management and union delegates/employee representatives—from all three of the participating organisations. For the impact and outcome evaluation studies, the population included all employees from the three participating organisations, and participants were all employees who consented to participate in the study and who completed both the pre-and post-policy implementation questionnaires. Qualitative methods in the form of interviews with key stakeholders were used to evaluate the process of developing and implementing the workplace AOD policy. In order to evaluate the impact of the policy with regard to the risk factors for workplace AOD impairment, and the outcome of the policy in terms of reducing workplace AOD impairment, quantitative methods in the form of a non-randomised single group pre- and post-test design were used. Changes from Time 1 (pre) to Time 2 (post) in the risk factors for workplace AOD impairment, and changes in the behaviour of interest—(self-reported) workplace AOD impairment—were measured. An integration of the findings from the process, impact, and outcome evaluation studies was undertaken using a combination of qualitative and quantitative methods. RESULTS For the process evaluation study Study respondents indicated that their policy was developed in the context of comparable industries across Australia developing workplace AOD policies, and that this was mainly out of concern for the deleterious health and safety impacts of workplace AOD impairment. Results from the process evaluation study also indicated that in developing and implementing the workplace AOD policy, there were mainly ‗winners', in terms of health and safety in the workplace. While there were some components of the development and implementation of the policy that were better done than others, and the process was expensive and took a long time, there were, overall, few unanticipated consequences to implementing the policy and it was reported to be thorough and of a high standard. Findings also indicated that overall the policy was developed and implemented according to best-practice in that: consultation during the policy development phase (with all the main stakeholders) was extensive; the policy was comprehensive; there was universal application of the policy to all employees; changes in the workplace (with regard to the policy) were gradual; and, the policy was publicised appropriately. Furthermore, study participants' responses indicated that the role of an independent external expert, who was trusted by all stakeholders, was integral to the success of the policy. For the impact and outcome evaluation studies Notwithstanding the limitations of pre- and post-test study designs with regard to attributing cause to the intervention, the findings from the impact evaluation study indicated that following policy implementation, statistically significant positive changes with regard to workplace AOD impairment were recorded for the following variables (risk factors for workplace AOD impairment): Knowledge; Attitudes; Perceived Behavioural Control; Perceptions of the Certainty of being punished for coming to work impaired by AODs; Perceptions of the Swiftness of punishment for coming to work impaired by AODs; and Direct and Indirect Experience with Punishment Avoidance for workplace AOD impairment. There were, however, no statistically significant positive changes following policy implementation for Behavioural Intentions, Subjective Norms, and Perceptions of the Severity of punishment for workplace AOD impairment. With regard to the outcome evaluation, there was a statistically significant reduction in self-reported workplace AOD impairment following the implementation of the policy. As with the impact evaluation, these findings need to be interpreted in light of the limitations of the study design in being able to attribute cause to the intervention alone. The findings from the outcome evaluation study also showed that while a positive change in self-reported workplace AOD impairment following implementation of the policy did not appear to be related to gender, age group, or employment type, it did appear to be related to levels of employee general alcohol use, cannabis use, site type, and employment role. Integration of the process, impact, and outcome evaluation studies There appeared to be qualitative support for the relationship between the process of developing and implementing the policy, and the impact of the policy in changing the risk factors for workplace AOD impairment. That is, overall the workplace AOD policy was developed and implemented well and, following its implementation, there were positive changes in the majority of measured risk factors for workplace AOD impairment. Quantitative findings lend further support for a relationship between the process and impact of the policy, in that there was a statistically significant association between employee perceived fidelity of the policy (related to the process of the policy) and positive changes in some risk factors for workplace AOD impairment (representing the impact of the policy). Findings also indicated support for the relationship between the impact of the policy in changing the risk factors for workplace AOD impairment and the outcome of the policy in reducing workplace AOD impairment: positive changes in the risk factors for workplace AOD impairment (impact) were related to positive changes in self reported workplace AOD impairment (representing the main goal and outcome of the policy). CONCLUSIONS The findings from the research indicate support for the conclusion that the policy was appropriately implemented and that it achieved its objectives and main goal. The Doctoral research findings also addressed a number of gaps in the literature on workplace AOD impairment, namely: the likely effectiveness of AOD policies for reducing AOD impairment in the workplace, which factors in the development and implementation of a workplace AOD policy are likely to facilitate or impede the effectiveness of the policy to reduce workplace AOD impairment, and which employee groups are less likely to respond well to policies of this type. The findings from this research not only represent an example of translational, applied research—through the evaluation of the study industry's policy—but also add to the body of knowledge on workplace AOD policies and provide policy-makers with evidence which may be useful in the development and implementation of effective workplace AOD policies. Importantly, the findings espouse the importance of scientific evidence in the development, implementation, and evaluation of workplace AOD policies.

R1STM: One-class support tensor machine with randomised kernel

Identifying unusual or anomalous patterns in an underlying dataset is an important but challenging task in many applications. The focus of the unsupervised anomaly detection literature has mostly been on vectorised data. However, many applications are more naturally described using higher-order tensor representations. Approaches that vectorise tensorial data can destroy the structural information encoded in the high-dimensional space, and lead to the problem of the curse of dimensionality. In this paper we present the first unsupervised tensorial anomaly detection method, along with a randomised version of our method. Our anomaly detection method, the One-class Support Tensor Machine (1STM), is a generalisation of conventional one-class Support Vector Machines to higher-order spaces. 1STM preserves the multiway structure of tensor data, while achieving significant improvement in accuracy and efficiency over conventional vectorised methods. We then leverage the theory of nonlinear random projections to propose the Randomised 1STM (R1STM). Our empirical analysis on several real and synthetic datasets shows that our R1STM algorithm delivers comparable or better accuracy to a state-of-the-art deep learning method and traditional kernelised approaches for anomaly detection, while being approximately 100 times faster in training and testing.

Design of multiunit electronic exchanges through decomposition

In this paper, we exploit the idea of decomposition to match buyers and sellers in an electronic exchange for trading large volumes of homogeneous goods, where the buyers and sellers specify marginal-decreasing piecewise constant price curves to capture volume discounts. Such exchanges are relevant for automated trading in many e-business applications. The problem of determining winners and Vickrey prices in such exchanges is known to have a worst-case complexity equal to that of as many as (1 + m + n) NP-hard problems, where m is the number of buyers and n is the number of sellers. Our method proposes the overall exchange problem to be solved as two separate and simpler problems: 1) forward auction and 2) reverse auction, which turns out to be generalized knapsack problems. In the proposed approach, we first determine the quantity of units to be traded between the sellers and the buyers using fast heuristics developed by us. Next, we solve a forward auction and a reverse auction using fully polynomial time approximation schemes available in the literature. The proposed approach has worst-case polynomial time complexity. and our experimentation shows that the approach produces good quality solutions to the problem. Note to Practitioners- In recent times, electronic marketplaces have provided an efficient way for businesses and consumers to trade goods and services. The use of innovative mechanisms and algorithms has made it possible to improve the efficiency of electronic marketplaces by enabling optimization of revenues for the marketplace and of utilities for the buyers and sellers. In this paper, we look at single-item, multiunit electronic exchanges. These are electronic marketplaces where buyers submit bids and sellers ask for multiple units of a single item. We allow buyers and sellers to specify volume discounts using suitable functions. Such exchanges are relevant for high-volume business-to-business trading of standard products, such as silicon wafers, very large-scale integrated chips, desktops, telecommunications equipment, commoditized goods, etc. The problem of determining winners and prices in such exchanges is known to involve solving many NP-hard problems. Our paper exploits the familiar idea of decomposition, uses certain algorithms from the literature, and develops two fast heuristics to solve the problem in a near optimal way in worst-case polynomial time.

Parametrized actor-critic algorithms for finite-horizon MDPs

Due to their non-stationarity, finite-horizon Markov decision processes (FH-MDPs) have one probability transition matrix per stage. Thus the curse of dimensionality affects FH-MDPs more severely than infinite-horizon MDPs. We propose two parametrized 'actor-critic' algorithms to compute optimal policies for FH-MDPs. Both algorithms use the two-timescale stochastic approximation technique, thus simultaneously performing gradient search in the parametrized policy space (the 'actor') on a slower timescale and learning the policy gradient (the 'critic') via a faster recursion. This is in contrast to methods where critic recursions learn the cost-to-go proper. We show w.p 1 convergence to a set with the necessary condition for constrained optima. The proposed parameterization is for FHMDPs with compact action sets, although certain exceptions can be handled. Further, a third algorithm for stochastic control of stopping time processes is presented. We explain why current policy evaluation methods do not work as critic to the proposed actor recursion. Simulation results from flow-control in communication networks attest to the performance advantages of all three algorithms.

Pengarna eller livet? : En kvalitativ och longitudinell studie om långtidsarbetslösa unga i ett aktörsperspektiv

Your money or your life? A qualitative follow-up study of the young unemployed from an actor perspective is a qualitative and longitudinal study following 36 unemployed young people in Helsinki over a span of ten years. The purpose of the study is to shed light on how a few young people view employment/unemployment and their lives and future, how they as unemployed perceive their encounters with society, and how society supports them. Four so-called key informants were followed at a finer level of empirical detail. They were chosen for the thematic interviews because of their different personalities, starting points and preferences. Although some differences were expected, what the results show is quite striking. The individual stories raise a number of questions about differences between young people, about society s view of the young unemployed, and about the principles behind the so-called activation policy and how society s support is distributed. The key informants descriptions underline that the group young unemployed does not consist of individuals who are alike but that life is complex, that paid work and unemployment can be perceived very differently, and that background and unofficial support can have consequences for self-perception and for ways of looking at the future, vocational choices, paid work and activation policy. Margaret S. Archer s theory of Morphogenesis and Barbara Cruikshank s theory of constructing democracies compose the study s theoretical framework. The key informants stories give a picture of a formal support system that, even though it puts part of the responsibility for unemployment on the individuals themselves, in the name of fairness and equality, treats them in an impersonal way, not giving their personal situation and wishes much weight. As a consequence, those who share the dominant values of society do well, while others who do not are faced with difficulties. The bigger the gap between society s and the individual s values, the bigger the risk to be met by little understanding and by penalties. And vice versa: Those who initially have the right values and know how to deal with authorities get heard and their opinions get accepted. The informants ask for a more personal encounter, which could improve both the atmosphere and the clients experiences of being heard. Still the risk of having a more individualistic system should be addressed, as a new system might generate new winners, but just as well give new losers. Finally, we have to ask if the so-called activation policy is looking for answers primarily to a macro-level problem on the micro-level. If it does not produce more jobs, its support for the unemployed will be insignificant. It is not enough to think about what to do at the grassroots level to make the system more functional and support job-seeking. If the current rate of unemployment endures, the quality of life of the unemployed should be addressed. A first step could be taken by placing less guilt on the unemployed. Instead of talking about activating the unemployed, discussion should be targeted at removing structural impediments to employment. If we want to have less polarisation between the those with paid work and those without, who often struggle with low incomes, we need to include the macro-level in the discussion. What does high unemployment mean in a work-based society, where the individual s self-perception and important social forms of support are linked to labour income? And what can be done at the macro-level to change this undesirable condition at the micro-level? Keywords: Unemployment, Youth, Public interventions, Activation policy, Individual actors, Qualitative, Longitudinal, Holistic, Helsinki, Finland

Kiista, eheys, unohdus : Vapaussodan muistaminen suojeluskunnan ja veteraaniliikkeen toiminnassa 1918-1944

Conflict, Unity, Oblivion: Commemoration of the Liberation War by the Civic Guard and the Veterans´ Union in 1918-1944 The Finnish Civil War ended in May 1918 as a victory for the white side. The war was named by the winners as the Liberation War and its legacy became a central theme for public commemorations during the interwar period. At the same time the experiences of the defeated were hindered from becoming a part of the official history of Finland. The commemoration of the war was related not only to the war experience but also to a national mission, which was seen fulfilled with the independence of Finland. Although the idea of the commemoration was to form a unifying non-political scene for the nation, the remembrance of the Liberation War rather continued than sought to reconcile to the conflict of 1918. The outbreak of the war between the Soviet Union and Finland in 1939 immediately affected the memory culture. The new myth of the Miracle of the Winter War, which referred to the unity shown by the people, required a marginalization of controversial memory of the Liberation War. This study examines from the concepts of public memory and narrative templates how the problematic experience of a civil war developed to a popular public commemoration. Instead of dealing with the manipulative and elite-centered grandiose commemoration projects, the study focuses on the more modest local level and emphasizes the significance of local memory agents and narrative templates of collective memory. The main subjects in the study are the Civil Guard and the Veterans´ Union. Essential for the widespread movement was the development of the Civic Guard from a wartime organization to a peacetime popular movement. The guards, who identified themselves trough the memories and the threats of civil war, formed a huge network of memory agents in every corner of the country. They effectively linked both local memory with official memory and the civic society with the state level. Only with the emergence of the right wing veteran movement in the 30ies did the tensions grow between the two levels of public memory. The study shows the diversity of the commemoration movement of the Liberation War. It was not only a result of a nation-state project and political propaganda, but also a way for local communities to identify and strengthen themselves in a time of political upheaval and uncertainty.

Kalman Filter Algorithm for Rating and Prediction in Basketball

The Thesis presents a state-space model for a basketball league and a Kalman filter algorithm for the estimation of the state of the league. In the state-space model, each of the basketball teams is associated with a rating that represents its strength compared to the other teams. The ratings are assumed to evolve in time following a stochastic process with independent Gaussian increments. The estimation of the team ratings is based on the observed game scores that are assumed to depend linearly on the true strengths of the teams and independent Gaussian noise. The team ratings are estimated using a recursive Kalman filter algorithm that produces least squares optimal estimates for the team strengths and predictions for the scores of the future games. Additionally, if the Gaussianity assumption holds, the predictions given by the Kalman filter maximize the likelihood of the observed scores. The team ratings allow probabilistic inference about the ranking of the teams and their relative strengths as well as about the teams’ winning probabilities in future games. The predictions about the winners of the games are correct 65-70% of the time. The team ratings explain 16% of the random variation observed in the game scores. Furthermore, the winning probabilities given by the model are concurrent with the observed scores. The state-space model includes four independent parameters that involve the variances of noise terms and the home court advantage observed in the scores. The Thesis presents the estimation of these parameters using the maximum likelihood method as well as using other techniques. The Thesis also gives various example analyses related to the American professional basketball league, i.e., National Basketball Association (NBA), and regular seasons played in year 2005 through 2010. Additionally, the season 2009-2010 is discussed in full detail, including the playoffs.

Vapaussodan perintö : Valkoisuus paikallisyhteisössä sisältö, ylläpitäjät ja rapautuminen Iisalmessa 1918 1933

Legacy of the Finnish Civil War. White nationalism in a local community - content, supporters and disintegration in Iisalmi 1918 - 1933. Using one local community (Iisalmi) as an example, this study centres around the winners of the 1918 Finnish Civil War, exploring their collectivity its subsequent breakdown during 1918 - 1933. Referring to this collectivity by the methodological concept of white nationalism, the thesis first discusses its origin, content and forms. This is done by elucidating the discourses and symbols that came to constitute central ideological and ritualistic elements of white nationalism. Next, the thesis describes and analyzes fundamental actors of the Finnish civil society (such as White Guard and Lotta Svärd) that maintained white nationalism as a form of counter or parallel hegemony to the integration policy of the 1920s. Also highlighted is the significance of white nationalism as a power broker and an instrument of moral regulation in inter-war Finnish society. A third contribution of this thesis involves presenting a new interpretation of the legacy of the Civil War, i.e., the right-wing radicalism during the years 1919 - 1933. I shall describe attempts of the extreme right (Lapua Movement and IKL, Patriotic People s Movement) to use the white nationalism discourse as a vehicle for their political ambitions, as well as the strong counter-reaction these attempts induced among other middle-class groups. At the core of this research is the concept of white nationalism, whose key elements were the sacrifice of 1918, fatherland under threat and warrior citizenship. Winners of the civil war strove to blend these ideals into a homogenized culture, to which the working class and wavering members of the middle-class were coaxed and pressurized to subscribe. The thesis draws on Anglo-American symbol theories, theory of social identity groups, Antonio Gramsci s concept of cultural hegemony and Stuart Hall s approach to discourse and power.

A Feature-Based Solution to Forward Problem in Electrical Capacitance Tomography of Conductive Materials

A new feature-based technique is introduced to solve the nonlinear forward problem (FP) of the electrical capacitance tomography with the target application of monitoring the metal fill profile in the lost foam casting process. The new technique is based on combining a linear solution to the FP and a correction factor (CF). The CF is estimated using an artificial neural network (ANN) trained using key features extracted from the metal distribution. The CF adjusts the linear solution of the FP to account for the nonlinear effects caused by the shielding effects of the metal. This approach shows promising results and avoids the curse of dimensionality through the use of features and not the actual metal distribution to train the ANN. The ANN is trained using nine features extracted from the metal distributions as input. The expected sensors readings are generated using ANSYS software. The performance of the ANN for the training and testing data was satisfactory, with an average root-mean-square error equal to 2.2%.

Reinforcement Learning With Function Approximation for Traffic Signal Control

We propose, for the first time, a reinforcement learning (RL) algorithm with function approximation for traffic signal control. Our algorithm incorporates state-action features and is easily implementable in high-dimensional settings. Prior work, e. g., the work of Abdulhai et al., on the application of RL to traffic signal control requires full-state representations and cannot be implemented, even in moderate-sized road networks, because the computational complexity exponentially grows in the numbers of lanes and junctions. We tackle this problem of the curse of dimensionality by effectively using feature-based state representations that use a broad characterization of the level of congestion as low, medium, or high. One advantage of our algorithm is that, unlike prior work based on RL, it does not require precise information on queue lengths and elapsed times at each lane but instead works with the aforementioned described features. The number of features that our algorithm requires is linear to the number of signaled lanes, thereby leading to several orders of magnitude reduction in the computational complexity. We perform implementations of our algorithm on various settings and show performance comparisons with other algorithms in the literature, including the works of Abdulhai et al. and Cools et al., as well as the fixed-timing and the longest queue algorithms. For comparison, we also develop an RL algorithm that uses full-state representation and incorporates prioritization of traffic, unlike the work of Abdulhai et al. We observe that our algorithm outperforms all the other algorithms on all the road network settings that we consider.

Hybrid approaches for clustering

Applications in various domains often lead to very large and frequently high-dimensional data. Successful algorithms must avoid the curse of dimensionality but at the same time should be computationally efficient. Finding useful patterns in large datasets has attracted considerable interest recently. The primary goal of the paper is to implement an efficient Hybrid Tree based clustering method based on CF-Tree and KD-Tree, and combine the clustering methods with KNN-Classification. The implementation of the algorithm involves many issues like good accuracy, less space and less time. We will evaluate the time and space efficiency, data input order sensitivity, and clustering quality through several experiments.

Generative Maximum Entropy Learning for Multiclass Classification

Maximum entropy approach to classification is very well studied in applied statistics and machine learning and almost all the methods that exists in literature are discriminative in nature. In this paper, we introduce a maximum entropy classification method with feature selection for large dimensional data such as text datasets that is generative in nature. To tackle the curse of dimensionality of large data sets, we employ conditional independence assumption (Naive Bayes) and we perform feature selection simultaneously, by enforcing a `maximum discrimination' between estimated class conditional densities. For two class problems, in the proposed method, we use Jeffreys (J) divergence to discriminate the class conditional densities. To extend our method to the multi-class case, we propose a completely new approach by considering a multi-distribution divergence: we replace Jeffreys divergence by Jensen-Shannon (JS) divergence to discriminate conditional densities of multiple classes. In order to reduce computational complexity, we employ a modified Jensen-Shannon divergence (JS(GM)), based on AM-GM inequality. We show that the resulting divergence is a natural generalization of Jeffreys divergence to a multiple distributions case. As far as the theoretical justifications are concerned we show that when one intends to select the best features in a generative maximum entropy approach, maximum discrimination using J-divergence emerges naturally in binary classification. Performance and comparative study of the proposed algorithms have been demonstrated on large dimensional text and gene expression datasets that show our methods scale up very well with large dimensional datasets.

Two timescale convergent Q-learning for sleep-scheduling in wireless sensor networks

In this paper, we consider an intrusion detection application for Wireless Sensor Networks. We study the problem of scheduling the sleep times of the individual sensors, where the objective is to maximize the network lifetime while keeping the tracking error to a minimum. We formulate this problem as a partially-observable Markov decision process (POMDP) with continuous stateaction spaces, in a manner similar to Fuemmeler and Veeravalli (IEEE Trans Signal Process 56(5), 2091-2101, 2008). However, unlike their formulation, we consider infinite horizon discounted and average cost objectives as performance criteria. For each criterion, we propose a convergent on-policy Q-learning algorithm that operates on two timescales, while employing function approximation. Feature-based representations and function approximation is necessary to handle the curse of dimensionality associated with the underlying POMDP. Our proposed algorithm incorporates a policy gradient update using a one-simulation simultaneous perturbation stochastic approximation estimate on the faster timescale, while the Q-value parameter (arising from a linear function approximation architecture for the Q-values) is updated in an on-policy temporal difference algorithm-like fashion on the slower timescale. The feature selection scheme employed in each of our algorithms manages the energy and tracking components in a manner that assists the search for the optimal sleep-scheduling policy. For the sake of comparison, in both discounted and average settings, we also develop a function approximation analogue of the Q-learning algorithm. This algorithm, unlike the two-timescale variant, does not possess theoretical convergence guarantees. Finally, we also adapt our algorithms to include a stochastic iterative estimation scheme for the intruder's mobility model and this is useful in settings where the latter is not known. Our simulation results on a synthetic 2-dimensional network setting suggest that our algorithms result in better tracking accuracy at the cost of only a few additional sensors, in comparison to a recent prior work.