Review of several stochastic speech unit models

Hidden Markov models are commonly used for speech unit modelling. This type of model is composed of a non-observable or “hidden” process, representing the temporal structure of the speech unit, and an observation process linking the hidden process with the acoustic parameters extracted from the speech signal.Different types of hidden processes (Markov chain, semi-Markov chain, “expanded-state” Markov chain) as well as different types of observation processes (discrete, continuous, semi-continuous—multiple processes) are reviewed, showing their relationships. The maximum likelihood estimation of two-stage stochastic process parameters is presented in an a posteriori probability formalism. An intepretation of the expectation-maximization algorithm is proposed and the practical learning algorithms for hidden Markov models and hidden semi-Markov models are compared in terms of computation structure, probabilistic justification and complexity.This presentation is illustrated by experiments on a multi-speaker 130 isolated word recognition system. The implementation techniques are detailed and the different combinations of state occupancy modelling techniques and observation modelling techniques are studied from a practical point of view.     

Statistical prioritization for software product line testing: an experience report

Software product lines (SPLs) are families of software systems sharing common assets and exhibiting variabilities specific to each product member of the family. Commonalities and variabilities are often represented as features organized in a feature model. Due to combinatorial explosion of the number of products induced by possible features combinations, exhaustive testing of SPLs is intractable. Therefore, sampling and prioritization techniques have been proposed to generate sorted lists of products based on coverage criteria or weights assigned to features. Solely based on the feature model, these techniques do not take into account behavioural usage of such products as a source of prioritization. In this paper, we assess the feasibility of integrating usage models into the testing process to derive statistical testing approaches for SPLs. Usage models are given as Markov chains, enabling prioritization of probable/rare behaviours. We used featured transition systems, compactly modelling variability and behaviour for SPLs, to determine which products are realizing prioritized behaviours. Statistical prioritization can achieve a significant reduction in the state space, and modelling efforts can be rewarded by better automation. In particular, we used MaTeLo, a statistical test cases generation suite developed at ALL4TEC. We assess feasibility criteria on two systems: Claroline, a configurable course management system, and Sferion?, an embedded system providing helicopter landing assistance.     

Layered shape grammars

In this article we propose a computer-aided conceptual design system to assist modelling at the early stages of design. More precisely, we address the problem of providing the designer with design alternatives that can be used as starting points of the design process. To guide the generation of such alternatives according to a given set of design requirements, the designer can express both visual knowledge in the form of basic geometric transformation rules, and also logic constraints that guide the modelling process. Our approach is based on the formalism of shape grammars, and supplements the basic algorithms with procedures that integrate logic design constraints and goals. Additionally, we introduce a layered scheme for shape grammars that can greatly reduce the computational cost of shape generation. Shape grammars, constraints, goals and layers can be handled through a graphic environment. We illustrate the functionalities of ShaDe through two use cases taken from the architectural design and video games domains, and also evaluate the performance of the system.     

A modelling and simulation based process for dependable systems design

Complex real-time system design needs to address dependability requirements, such as safety, reliability, and security. We introduce a modelling and simulation based approach which allows for the analysis and prediction of dependability constraints. Dependability can be improved by making use of fault tolerance techniques. The de-facto example, in the real-time system literature, of a pump control system in a mining environment is used to demonstrate our model-based approach. In particular, the system is modelled using the Discrete EVent system Specification (DEVS) formalism, and then extended to incorporate fault tolerance mechanisms. The modularity of the DEVS formalism facilitates this extension. The simulation demonstrates that the employed fault tolerance techniques are effective. That is, the system performs satisfactorily despite the presence of faults. This approach also makes it possible to make an informed choice between different fault tolerance techniques. Performance metrics are used to measure the reliability and safety of the system, and to evaluate the dependability achieved by the design. In our model-based development process, modelling, simulation and eventual deployment of the system are seamlessly integrated.     

Beyond Tracking: Modelling Activity and Understanding Behaviour

In this work, we present a unified bottom-up and top-down automatic model selection based approach for modelling complex activities of multiple objects in cluttered scenes. An activity of multiple objects is represented based on discrete scene events and their behaviours are modelled by reasoning about the temporal and causal correlations among different events. This is significantly different from the majority of the existing techniques that are centred on object tracking followed by trajectory matching. In our approach, object-independent events are detected and classified by unsupervised clustering using Expectation-Maximisation (EM) and classified using automatic model selection based on Schwarz's Bayesian Information Criterion (BIC). Dynamic Probabilistic Networks (DPNs) are formulated for modelling the temporal and causal correlations among discrete events for robust and holistic scene-level behaviour interpretation. In particular, we developed a Dynamically Multi-Linked Hidden Markov Model (DML-HMM) based on the discovery of salient dynamic interlinks among multiple temporal processes corresponding to multiple event classes. A DML-HMM is built using BIC based factorisation resulting in its topology being intrinsically determined by the underlying causality and temporal order among events. Extensive experiments are conducted on modelling activities captured in different indoor and outdoor scenes. Our experimental results demonstrate that the performance of a DML-HMM on modelling group activities in a noisy and cluttered scene is superior compared to those of other comparable dynamic probabilistic networks including a Multi-Observation Hidden Markov Model (MOHMM), a Parallel Hidden Markov Model (PaHMM) and a Coupled Hidden Markov Model (CHMM). First online version published in February, 2006     

Variational Surface Design and Surface Interrogation

The generation of technical smooth surfaces from a mesh of three-dimensional data points is an important problem in geometric modelling. In this publication we give a survey of some new techniques based on a calculus of variation approach. Apart from the pure construction of these surfaces, the analysis of their quality is equally important in the design and manufacturing process. Generalized focal surfaces are presented here as a new surface interrogation tool.     

Markovian modelling and Fisher distribution for unsupervised classification of radar images

Statistical segmentation techniques based on hidden Markov field modelling have generated considerable interest in past years. They take contextual information into account in a particularly elegant and rigorous way. Although these models have been thoroughly tested, they can fail in some cases such as the non-stationary one. In this article, we propose use of the recently developed triplet Markov field, which models non-stationary images, and that of Fisher distribution, which is adapted to a wide range of surfaces for modelling synthetic aperture radar (SAR) image noise. Examples illustrate the difference between the approach proposed and classical ones. Various experiments indicate that the new model and its associated unsupervised algorithm perform better than classical ones.     

Meta-modelling and graph grammars for multi-paradigm modelling in AToM<Superscript>3</Superscript>

This paper presents the combined use of meta-modelling and graph grammars for the generation of visual modelling tools for simulation formalisms. In meta-modelling, formalisms are described at a meta-level. This information is used by a meta-model processor to generate modelling tools for the described formalisms. We combine meta-modelling with graph grammars to extend the model manipulation capabilities of the generated modelling tools: edit, simulate, transform into another formalism, optimize and generate code. We store all (meta-)models as graphs, and thus, express model manipulations as graph grammars.We present the design and implementation of these concepts in AToM³ (A_To_ol for M_ulti-formalism, M_eta-M_odelling). AToM³ supports modelling of complex systems using different formalisms, all meta-modelled in their own right. Models in different formalisms may be transformed into a single common formalism for further processing. These transformations are specified by graph grammars. Mosterman and Vangheluwe [18] introduced the term multi-paradigm modelling to denote the combination of multiple formalisms, multiple abstraction levels, and meta-modelling. As an example of multi-paradigm modelling we present a meta-model for the Object-Oriented Continuous Simulation Language OOCSMP, in which we combine ideas from UML class diagrams (to express the OOCSMP model structure), Causal Block Diagrams (CBDs), and Statecharts (to specify the methods of the OOCSMP classes). A graph grammar is able to generate OOCSMP code, and then a compiler for this language (C-OOL) generates Java applets for the simulation execution.     

An illustration of verification and validation in the modelling phase of KBS development

Reliability has become a key factor in KBS development. For this reason, it has been suggested that verification and validation (V&V) should become an integrated part of activities throughout the whole KBS development cycle. In this paper, it will be illustrated how the PROLOGA workbench integrates V&V aspects into its modelling environment, such that these techniques can be of assistance in the process of knowledge acquisition and representation. To this end, verification has to be performed incrementally and can no longer be delayed until after the system has been completed. It will be shown how this objective can be realised through an approach that uses the decision table formalism as a modelling instrument.     

Requirements Engineering-Based Conceptual Modelling

The software production process involves a set of phases where a clear relationship and smooth transitions between them should be introduced. In this paper, a requirements engineering-based conceptual modelling approach is introduced as a way to improve the quality of the software production process. The aim of this approach is to provide a set of techniques and methods to capture software requirements and to provide a way to move from requirements to a conceptual schema in a traceable way. The approach combines a framework for requirements engineering (TRADE) and a graphical object-oriented method for conceptual modelling and code generation (OO-Method). The intended improvement of the software production process is accomplished by providing a precise methodological guidance to go from the user requirements (represented through the use of the appropriate TRADE techniques) to the conceptual schema that properly represents them (according to the conceptual constructs provided by the OO-Method). Additionally, as the OO-Method provides full model-based code generation features, this combination minimises the time dedicated to obtaining the final software product.     

基于强化学习算法的神经网络模糊测试技术优化研究

现有神经网络模糊测试技术在测试样本生成阶段通常对初始样本进行随机变异,导致生成样本质量不高,从而测试覆盖率不高;针对以上问题,提出一种基于强化学习算法的神经网络模糊测试技术,将模糊测试过程建模为马尔可夫决策过程,在该模型中,测试样本被看作环境状态,不同的变异方法被看作可供选择的动作空间,神经元覆盖率被看作奖励反馈,使用强化学习算法来学习最优的变异策略,指导生成最优测试样本,使其能够获得最高的神经元覆盖率;通过与现有的主流神经网络模糊测试方法的对比实验表明,基于强化学习算法的神经网络模糊测试技术,可以提升在不同粒度下的神经元覆盖。     

Generating multiple alternative clusterings via globally optimal subspaces

Clustering analysis is important for exploring complex datasets. Alternative clustering analysis is an emerging subfield involving techniques for the generation of multiple different clusterings, allowing the data to be viewed from different perspectives. We present two new algorithms for alternative clustering generation. A distinctive feature of our algorithms is their principled formulation of an objective function, facilitating the discovery of a subspace satisfying natural quality and orthogonality criteria. The first algorithm is a regularization of the Principal Components analysis method, whereas the second is a regularization of graph-based dimension reduction. In both cases, we demonstrate a globally optimum subspace solution can be computed. Experimental evaluation shows our techniques are able to equal or outperform a range of existing methods.     

SoftPM: a software process management system reconciling formalism with easiness

Various formal approaches to process modeling and analysis have been proposed. With the emerging importance of practicality in this field, easiness in adopting formal technology should be taken into account. In this paper, we propose a PSEE called SoftPM that is based on a high level Petri net formalism for process modeling. To overcome the difficulty in using this formalism, SoftPM exploits a multi-level modeling mechanism for the representation of software processes. SoftPM supports three different levels for process representation. They are cognitive, MAM net, and Pr/T net. Each level has different roles in SoftPM. The cognitive-level modeling provides the means of getting free from difficulties in manipulating the modeling formalism. MAM net takes the role of core modeling formalism in SoftPM. Finally, Pr/T nets support the low-level modeling formalism as an existing Petri-net class. Moreover, SoftPM offers various analysis techniques to aid managerial decision making, as well as conventional Petri-net analysis techniques. Using a Java-based thin client/server architecture, SoftPM smoothly supports execution at distributed heterogeneous platforms over the Internet.     

Design with shape grammars and reinforcement learning

Shape grammars are a powerful and appealing formalism for automatic shape generation in computer-based design systems. This paper presents a proposal complementing the generative power of shape grammars with reinforcement learning techniques. We use simple (naive) shape grammars capable of generating a large variety of different designs. In order to generate those designs that comply with given design requirements, the grammar is subject to a process of machine learning using reinforcement learning techniques. Based on this method, we have developed a system for architectural design, aimed at generating two-dimensional layout schemes of single-family housing units. Using relatively simple grammar rules, we learn to generate schemes that satisfy a set of requirements stated in a design guideline. Obtained results are presented and discussed.     

Modelling of batch production systems using Petri nets with dynamic tokens

This article deals with the qualitative modelling of batch production plants. The approach held is Petri net based. First a definition of a three-level net formalism is presented; the formalism extends the Valk's approach of nets in nets in which the tokens are other nets; in this extension the lower level net handles symbolic tokens. Then a methodology for the modelling of batch processes is proposed; in this methodology the upper level describes the plant layout, the next level models an entity that goes with the batch through the plant and specifies the material and the process routes that the batch must follow, and the third level represents detailed treatments to perform into the process cells using specific equipment. The use of the modelling formalism is illustrated through an example dealing with the coordination of a batch manufacturing system.     

Intelligent support for systems development: The notion and the issues

Enterprise modelling and information systems modelling have traditionally utilized techniques developed in the earlier disciplines of systems analysis and operational analysis. However, these tools have proved insufficient even for information systems modelling and their inadequacies make them less than ideal for enterprise modelling. Furthermore, it has proved difficult to integrate the techniques into a uniform framework representation. Extensive research since the early 1980s has produced support tools for information systems engineering in the academic sector which have generally failed to reach widespread commercial use. Commercially developed integrated support systems aimed at enterprise modelling and information system modelling use traditional techniques, with a lack of formalism. The purpose of this paper is to introduce a technique which overcomes the major inadequacies and which provides an integrating framework to represent both the information and the processing, at enterprise and systems modelling levels of abstraction. The technique is based on the use of a transition network, but extended to represent enterprise and system models in a meaningful way. The use of an intelligent repository, with associated processing of the formally defined requirements, specification and design statements, is novel, and provides the additional support to make the technique amenable to handling the design process from informal to formal specification. The use of a graphical user interface, linked directly to the deductive system and repository, ensures intuitive ease of use.     

Predicting WWW surfing using multiple evidence combination

The improvement of many applications such as web search, latency reduction, and personalization/ recommendation systems depends on surfing prediction. Predicting user surfing paths involves tradeoffs between model complexity and predictive accuracy. In this paper, we combine two classification techniques, namely, the Markov model and Support Vector Machines (SVM), to resolve prediction using Dempster’s rule. Such fusion overcomes the inability of the Markov model in predicting the unseen data as well as overcoming the problem of multiclassification in the case of SVM, especially when dealing with large number of classes. We apply feature extraction to increase the power of discrimination of SVM. In addition, during prediction we employ domain knowledge to reduce the number of classifiers for the improvement of accuracy and the reduction of prediction time. We demonstrate the effectiveness of our hybrid approach by comparing our results with widely used techniques, namely, SVM, the Markov model, and association rule mining.     

The initial conditions problem in longitudinal count process: A simulation study

In this paper we report some simulation studies to compare two basic approaches in modelling the random effects for longitudinal count data. The first, the naive approach, treats the endogenous initial conditions as exogenous and the second, correct approach, models the initial conditions as endogenous. The initial conditions problem in longitudinal count data arises when the process has a Markov property. In some applications, such as clinical studies, the baseline is used as a covariate. This also creates an initial condition problem even if the process has no Markov property. In this paper we shall consider both types of initial conditions. To see the effect of the error variance, baseline effect and lagged effect different values for these parameters are considered. We will apply the proposed approach to epileptic data and show that ignoring the initial conditions results in exaggerated treatment effect and misleading interpretation for the baseline effect.     

Towards a dynamical model for wireless sensor networks

In this paper we introduce two dynamical models for a broadcast process in wireless sensor networks. We obtain a convergent martingale sequence for the two models. To our knowledge, such martingales were unknown previously. We look at the formal models using the formalisms of martingales, dynamical systems and Markov chains, each formalism providing complementary and coherent information with each other. The dynamics of both models are comparable and are validated in their domain of application with numerical simulation of wireless sensor networks. We make explicit the situations where the models are realistic. We also provide a formal analysis of the quasi-stationary distribution associated to the Markov chain corresponding to the second model proposed.