Statistical study of talk spurts and silence lengths with relation to various languages and speakers for advanced silence suppression in Next Generation Wireless Communication Devices

As wireless network technologies evolve towards an All-IP framework, Next Generation Wireless Communication Devices demand better use of spectral resources by employing advanced techniques of silence suppression. This paper presents an analysis of VoIP call data and compares the statistical results based on observed patterns of talk spurts and silence lengths to those achieved by a modified on-off voice model for silence suppression in wireless networks. As talk spurts and silence lengths are sensitive to varying word lengths, temporal structure and other prosodic aspects of speech, the impact of the use of various languages, dialects and gender of speakers on these results is also assessed.

Noisy random Boolean formulae:a statistical physics perspective

Properties of computing Boolean circuits composed of noisy logical gates are studied using the statistical physics methodology. A formula-growth model that gives rise to random Boolean functions is mapped onto a spin system, which facilitates the study of their typical behavior in the presence of noise. Bounds on their performance, derived in the information theory literature for specific gates, are straightforwardly retrieved, generalized and identified as the corresponding macroscopic phase transitions. The framework is employed for deriving results on error-rates at various function-depths and function sensitivity, and their dependence on the gate-type and noise model used. These are difficult to obtain via the traditional methods used in this field.

An advanced real-time predictive maintenance framework for large scale machine systems

This thesis introduces and develops a novel real-time predictive maintenance system to estimate the machine system parameters using the motion current signature. Recently, motion current signature analysis has been addressed as an alternative to the use of sensors for monitoring internal faults of a motor. A maintenance system based upon the analysis of motion current signature avoids the need for the implementation and maintenance of expensive motion sensing technology. By developing nonlinear dynamical analysis for motion current signature, the research described in this thesis implements a novel real-time predictive maintenance system for current and future manufacturing machine systems. A crucial concept underpinning this project is that the motion current signature contains infor­mation relating to the machine system parameters and that this information can be extracted using nonlinear mapping techniques, such as neural networks. Towards this end, a proof of con­cept procedure is performed, which substantiates this concept. A simulation model, TuneLearn, is developed to simulate the large amount of training data required by the neural network ap­proach. Statistical validation and verification of the model is performed to ascertain confidence in the simulated motion current signature. Validation experiment concludes that, although, the simulation model generates a good macro-dynamical mapping of the motion current signature, it fails to accurately map the micro-dynamical structure due to the lack of knowledge regarding performance of higher order and nonlinear factors, such as backlash and compliance. Failure of the simulation model to determine the micro-dynamical structure suggests the pres­ence of nonlinearity in the motion current signature. This motivated us to perform surrogate data testing for nonlinearity in the motion current signature. Results confirm the presence of nonlinearity in the motion current signature, thereby, motivating the use of nonlinear tech­niques for further analysis. Outcomes of the experiment show that nonlinear noise reduction combined with the linear reverse algorithm offers precise machine system parameter estimation using the motion current signature for the implementation of the real-time predictive maintenance system. Finally, a linear reverse algorithm, BJEST, is developed and applied to the motion current signature to estimate the machine system parameters.

Statistical mechanics of distribution networks

This thesis presents an analysis of the stability of complex distribution networks. We present a stability analysis against cascading failures. We propose a spin [binary] model, based on concepts of statistical mechanics. We test macroscopic properties of distribution networks with respect to various topological structures and distributions of microparameters. The equilibrium properties of the systems are obtained in a statistical mechanics framework by application of the replica method. We demonstrate the validity of our approach by comparing it with Monte Carlo simulations. We analyse the network properties in terms of phase diagrams and found both qualitative and quantitative dependence of the network properties on the network structure and macroparameters. The structure of the phase diagrams points at the existence of phase transition and the presence of stable and metastable states in the system. We also present an analysis of robustness against overloading in the distribution networks. We propose a model that describes a distribution process in a network. The model incorporates the currents between any connected hubs in the network, local constraints in the form of Kirchoff's law and a global optimizational criterion. The flow of currents in the system is driven by the consumption. We study two principal types of model: infinite and finite link capacity. The key properties are the distributions of currents in the system. We again use a statistical mechanics framework to describe the currents in the system in terms of macroscopic parameters. In order to obtain observable properties we apply the replica method. We are able to assess the criticality of the level of demand with respect to the available resources and the architecture of the network. Furthermore, the parts of the system, where critical currents may emerge, can be identified. This, in turn, provides us with the characteristic description of the spread of the overloading in the systems.

A novel framework of training hidden Markov support vector machines from lightly-annotated data

Natural language understanding (NLU) aims to map sentences to their semantic mean representations. Statistical approaches to NLU normally require fully-annotated training data where each sentence is paired with its word-level semantic annotations. In this paper, we propose a novel learning framework which trains the Hidden Markov Support Vector Machines (HM-SVMs) without the use of expensive fully-annotated data. In particular, our learning approach takes as input a training set of sentences labeled with abstract semantic annotations encoding underlying embedded structural relations and automatically induces derivation rules that map sentences to their semantic meaning representations. The proposed approach has been tested on the DARPA Communicator Data and achieved 93.18% in F-measure, which outperforms the previously proposed approaches of training the hidden vector state model or conditional random fields from unaligned data, with a relative error reduction rate of 43.3% and 10.6% being achieved.

Semi-supervised learning of statistical models for natural language understanding

Natural language understanding is to specify a computational model that maps sentences to their semantic mean representation. In this paper, we propose a novel framework to train the statistical models without using expensive fully annotated data. In particular, the input of our framework is a set of sentences labeled with abstract semantic annotations. These annotations encode the underlying embedded semantic structural relations without explicit word/semantic tag alignment. The proposed framework can automatically induce derivation rules that map sentences to their semantic meaning representations. The learning framework is applied on two statistical models, the conditional random fields (CRFs) and the hidden Markov support vector machines (HM-SVMs). Our experimental results on the DARPA communicator data show that both CRFs and HM-SVMs outperform the baseline approach, previously proposed hidden vector state (HVS) model which is also trained on abstract semantic annotations. In addition, the proposed framework shows superior performance than two other baseline approaches, a hybrid framework combining HVS and HM-SVMs and discriminative training of HVS, with a relative error reduction rate of about 25% and 15% being achieved in F-measure.

A framework for automated construction of resource space based on background knowledge

Resource Space Model is a kind of data model which can effectively and flexibly manage the digital resources in cyber-physical system from multidimensional and hierarchical perspectives. This paper focuses on constructing resource space automatically. We propose a framework that organizes a set of digital resources according to different semantic dimensions combining human background knowledge in WordNet and Wikipedia. The construction process includes four steps: extracting candidate keywords, building semantic graphs, detecting semantic communities and generating resource space. An unsupervised statistical language topic model (i.e., Latent Dirichlet Allocation) is applied to extract candidate keywords of the facets. To better interpret meanings of the facets found by LDA, we map the keywords to Wikipedia concepts, calculate word relatedness using WordNet's noun synsets and construct corresponding semantic graphs. Moreover, semantic communities are identified by GN algorithm. After extracting candidate axes based on Wikipedia concept hierarchy, the final axes of resource space are sorted and picked out through three different ranking strategies. The experimental results demonstrate that the proposed framework can organize resources automatically and effectively.©2013 Published by Elsevier Ltd. All rights reserved.

Az extrém ármozgások statisztikai jellemzői a magyar áramtőzsdén (The statistical characteristics of extreme price movements in the Hungarian power exchange)

2010. július 20-án megkezdte működését a magyar áramtőzsde, a HUPX. 2010. augusztus 16-án az első napokban tapasztalt 45-60 euró megawattórás ár helyett egyes órákban 2999 eurós árral szembesültek a piaci szereplők. A kiemelkedően magas árak megjelenése nem szokatlan az áramtőzsdéken a nemzetközi tapasztalatok szerint, sőt a kutatások kiemelten foglalkoznak az ún. ártüskék okainak felkutatásával, valamint megjelenésük kvantitatív és kvalitatív elemzésével. A cikkben a szerző bemutatja, milyen eredmények születtek a kiugró árak statisztikai vizsgálatai során a szakirodalomban, illetve azok következtetései hogyan állják meg a helyüket a magyar árak idősorát figyelembe véve. A szerző bemutat egy modellkeretet, amely a villamosenergia-árak viselkedését a hét órái szerint periodikusan váltakozó paraméterű eloszlásokkal írja le. A magyar áramtőzsde rövid története sajnos nem teszi lehetővé, hogy a hét minden órájára külön áreloszlást illeszthessünk. A szerző ezért a hét óráit két csoportba sorolja az ár eloszlásának jellege alapján: az ártüskék megjelenése szempontjából kockázatos és kevésbé kockázatos órákba. Ezután a HUPX-árak leírására felépít egy determinisztikus, kétállapotú rezsimváltó modellt, amellyel azonosítani lehet a kockázatos és kevésbé kockázatos órákat, valamint képet kaphatunk az extrém ármozgások jellegéről. / === / On 20th July, 2010 the Hungarian Power Exchange, the HUPX started its operation. On 16th August in certain hours the markets participants faced € 2,999 price instead of in the first days experienced 45-60 euros/mwh. According to the international experiences the appearance of the extremely high prices hasn’t been unusual in the power exchanges, the researches have focused exploring the causes of the so-called spikes and quantitative and qualitative analysis of those appearances. In this article the author describes what results were determined on statistical studies of outstanding prices in the literature, and how their conclusions stand up into account the time series of the Hungarian prices. The author presents a model framework which describes the behavior of electricity prices in the seven hours of periodically varying parameters. Unfortunately the brief history of the Hungarian Power Exchange does not allow to suit specific prices for each hour of week. Therefore the author classifies the hours of the week in the two groups based on the nature of price dispersion: according to the appearance of spikes to risky and less risky classes. Then for describing the HUPX prices the author builds a deterministic two-state, regime-changing model, which can be identified the risky and less risky hours, and to get a picture of the nature of extreme price movements.

A diagnostic framework for demand amplification problems in supply chains

This dissertation delivers a framework to diagnose the Bull-Whip Effect (BWE) in supply chains and then identify methods to minimize it. Such a framework is needed because in spite of the significant amount of literature discussing the bull-whip effect, many companies continue to experience the wide variations in demand that are indicative of the bull-whip effect. While the theory and knowledge of the bull-whip effect is well established, there still is the lack of an engineering framework and method to systematically identify the problem, diagnose its causes, and identify remedies. ^ The present work seeks to fill this gap by providing a holistic, systems perspective to bull-whip identification and diagnosis. The framework employs the SCOR reference model to examine the supply chain processes with a baseline measure of demand amplification. Then, research of the supply chain structural and behavioral features is conducted by means of the system dynamics modeling method. ^ The contribution of the diagnostic framework, is called Demand Amplification Protocol (DAMP), relies not only on the improvement of existent methods but also contributes with original developments introduced to accomplish successful diagnosis. DAMP contributes a comprehensive methodology that captures the dynamic complexities of supply chain processes. The method also contributes a BWE measurement method that is suitable for actual supply chains because of its low data requirements, and introduces a BWE scorecard for relating established causes to a central BWE metric. In addition, the dissertation makes a methodological contribution to the analysis of system dynamic models with a technique for statistical screening called SS-Opt, which determines the inputs with the greatest impact on the bull-whip effect by means of perturbation analysis and subsequent multivariate optimization. The dissertation describes the implementation of the DAMP framework in an actual case study that exposes the approach, analysis, results and conclusions. The case study suggests a balanced solution between costs and demand amplification can better serve both firms and supply chain interests. Insights pinpoint to supplier network redesign, postponement in manufacturing operations and collaborative forecasting agreements with main distributors.^

Achieving zero accidents—A strategic framework for continuous safety improvement in the construction industry

In the U.S., construction accidents remain a significant economic and social problem. Despite recent improvement, the Construction industry, generally, has lagged behind other industries in implementing safety as a total management process for achieving zero accidents and developing a high-performance safety culture. One aspect of this total approach to safety that has frustrated the construction industry the most has been “measurement”, which involves identifying and quantifying the factors that critically influence safe work behaviors. The basic problem attributed is the difficulty in assessing what to measure and how to measure it—particularly the intangible aspects of safety. Without measurement, the notion of continuous improvement is hard to follow. This research was undertaken to develop a strategic framework for the measurement and continuous improvement of total safety in order to achieve and sustain the goal of zero accidents, while improving the quality, productivity and the competitiveness of the construction industry as it moves forward. The research based itself on an integral model of total safety that allowed decomposition of safety into interior and exterior characteristics using a multiattribute analysis technique. Statistical relationships between total safety dimensions and safety performance (measured by safe work behavior) were revealed through a series of latent variables (factors) that describe the total safety environment of a construction organization. A structural equation model (SEM) was estimated for the latent variables to quantify relationships among them and between these total safety determinants and safety performance of a construction organization. The developed SEM constituted a strategic framework for identifying, measuring, and continuously improving safety as a total concern for achieving and sustaining the goal of zero accidents.

The Olivia Framework:Concepts for Use in Finely-Grained, Integrated Social Policy Analysis (Working Paper 45)

The Olivia framework is a set of concepts and measures that, when mature, will allow users to describe, in a consistent and integrated manner, everything about individuals and institutions that is of potential interest to social policy. The present paper summarizes the current stage of development in achieving this highly ambitious goal. The current version of the framework supports analysis of social trends and policy responses from many perspectives: • The point-in-time, resource-flow perspectives that underlie most traditional, economics-based policy analysis. • Life-course perspectives, including both transitions/trajectories analysis and asset-based analysis. • Spatial perspectives that anchor people in space and history and that provide a link to macro-analysis. • The perspective of the purposes/goals of individuals and institutions, including the objectives of different types of government programming. The concepts of the framework, which are all potentially measurable, provide a language that can support integrated analysis in all these areas at a much finer level of description than is customary. It provides a language that is especially well suited for analysis of the incremental policy changes that are typical of a mature welfare state. It supports both qualitative and quantitative analysis, enabling some integration between the two. It supports citizen-centric as well as a government-centric view of social policy. In its current version, the concepts are most highly developed as they related to social policies as they related to labour markets, equality and social integration, care-giving, immigration, income security, sustainability, and social and economic well-being more generally. However the paper points to likely extensions in the areas of health, justice and safety.

Evaluating statistical and machine learning methods to predict risk of in-hospital child mortality in Uganda

Thesis (Master's)--University of Washington, 2016-08

Convex Optimization Algorithms and Statistical Bounds for Learning Structured Models

Thesis (Ph.D.)--University of Washington, 2016-08

Indoor environment and the impact on the health of pre-exisiting asthmatics at work: the development of a risk management framework

This thesis examines the spatial and temporal variation in nitrogen dioxide (NO2) levels in Guernsey and the impacts on pre-existing asthmatics. Whilst air quality in Guernsey is generally good, the levels of NO2 exceed UK standards in several locations. The evidence indicates that people suffering from asthma have exacerbation of their symptoms if exposed to elevated levels of air pollutants including NO2, although this research has never been carried out in Guernsey before. In addition, exposure assessment of individuals is rarely carried out and research in this area is limited due to the complexity of undertaking such a study, which will include a combination of exposures in the home, the workplace and ambient exposures, which vary depending on the individual daily experience. For the first time in Guernsey, this research has examined NO2 levels in correlation with asthma patient admissions to hospital, assessment of NO2 exposures in typical homes and typical workplaces in Guernsey. The data showed a temporal correlation between NO2 levels and the number of hospital admissions and the trend from 2008-2012 was upwards. Statistical analysis of the data did not show a significant linear correlation due to the small size of the data sets. Exposure assessment of individuals showed a spatial variation in exposures in Guernsey and assessment in indoor environments showed that real-time analysis of NO2 levels needs to be undertaken if indoor micro environments for NO2 are the be assessed adequately. There was temporal and spatial variation in NO2 concentrations measured using diffusion tubes, which provide a monthly mean value, and analysers measuring NO2 concentrations in real time. The research shows that building layout and design are important factors for good air flow and ventilation and the dispersion of NO2 indoors. Environmental Health Officers have statutory responsibilities for ambient air quality, hygiene of buildings and workplace environments and this role needs to be co-ordinated with healthcare professionals to improve health outcomes for asthmatics. The outcome of the thesis was the development of a risk management framework for pre-existing asthmatics at work for use by regulators of workplaces and an information leaflet to assist in improving health outcomes for asthmatics in Guernsey.

The renormalization group via statistical inference

In physics, one attempts to infer the rules governing a system given only the results of imperfect measurements. Hence, microscopic theories may be effectively indistinguishable experimentally. We develop an operationally motivated procedure to identify the corresponding equivalence classes of states, and argue that the renormalization group (RG) arises from the inherent ambiguities associated with the classes: one encounters flow parameters as, e.g., a regulator, a scale, or a measure of precision, which specify representatives in a given equivalence class. This provides a unifying framework and reveals the role played by information in renormalization. We validate this idea by showing that it justifies the use of low-momenta n-point functions as statistically relevant observables around a Gaussian hypothesis. These results enable the calculation of distinguishability in quantum field theory. Our methods also provide a way to extend renormalization techniques to effective models which are not based on the usual quantum-field formalism, and elucidates the relationships between various type of RG.