Pattern recognition by means of disjoint principal components models

Pattern recognition based on modelling each separate class by a separate principal components (PC) model is discussed. These PC models are shown to be able to approximate any continuous variation within a single class. Hence, methods based on PC models will, provided that the data are sufficient, recognize any pattern that exists in a given set of objects. In addition, fitting the objects in each class by a separate PC model will, in a simple way, provide information about such matters as the relevance of single variables, “outliers” among the objects and “distances” between different classes. Application to the classical Iris-data of Fisher is used as an illustration.     

基于神经网络编码的真值发现

由于互联网的开放性和多源性,不同互联网平台提供的数据参差不齐,多个数据源对同一实体的描述可能存在冲突,真值发现是消解语义冲突,提高数据质量的重要技术手段之一。传统真值发现算法通常假设数据源可靠度与观测值可信度间的关系可用简单函数表示,设计迭代规则或概率模型进行真值发现,而人工定义的条件通常难以反映数据底层的真实分布,导致真值发现结果不理想。针对此问题,提出基于神经网络编码的真值发现方法TDNNE。首先利用“数据源-数据源”“数据源-观测值”关系及真值发现的假设构造双损失深度神经网络;然后利用该网络将数据源与观测值嵌入到高维空间,分别表示数据源可靠度与观测值可信度,使可靠数据源与可信观测值彼此接近（同时,不可靠数据源与不可信观测值彼此接近）;最后基于嵌入空间进行真值发现。与传统方法相比,TDNNE方法不需要人工定义迭代规则或数据分布,而是利用神经网络自动学习数据源观测值间复杂的关系依赖。在真实数据集上的实验结果表明,该方法准确率较基于迭代的Accu等方法准确率提高约2%~25%,较基于概率图模型的3-Estimates等方法提高约2%~4%,较基于优化的CRH方法提高约2%~5%,较基于神经网络的FFMN方法提高约1%~2%。     

Complex agent networks: An emerging approach for modeling complex systems

Complexity and complex systems are all around us: from molecular and cellular systems in biology up to economics and human societies. There is an urgent need for methods that can capture the multi-scale spatio-temporal characteristics of complex systems. Recent emphasis has centered on two methods in particular, those being complex networks and agent-based models. In this paper we look at the combination of these two methods and identify “Complex Agent Networks”, as a new emerging computational paradigm for complex system modeling. We argue that complex agent networks are able to capture both individual-level dynamics as well as global-level properties of a complex system, and as such may help to obtain a better understanding of the fundamentals of such systems.     

Emotional agents: A modeling and an application

This paper proposes modeling of artificial emotions through agents based on symbolic approach. The symbolic approach utilizes symbolic emotional rule-based systems (rule base that generated emotions) with continuous interactions with environment and an internal “thinking” machinery that comes as a result of series of inferences, evaluation, evolution processes, adaptation, learning, and emotions. We build two models for agent based systems; one is supported with artificial emotions and the other one without emotions. We use both in solving a bench mark problem; “The Orphanage Care Problem”. The two systems are simulated and results are compared. Our study shows that systems with proper model of emotions can perform in many cases better than systems without emotions. We try to shed the light here on how artificial emotions can be modeled in a simple rule-based agent systems and if emotions as they exist in “real intelligence” can be helpful for “artificial intelligence”. Agent architectures are presented as a generic blueprint on which the design of agents can be based. Our focus is on the functional design, including flow of information and control. With this information provided, the generic blueprints of architectures should not be difficult to implement agents, thus putting these theoretical models into practice. We build the agents using this architecture, and many experiments and analysis are shown.     

FRASel: a consensus of feature ranking methods for time series modelling

The main objective of “time series analysis” is to discover the underlying structure of the time series, and thus, become able to forecast its “future values”. This process makes it possible to predict, control or simulate variables. Most of the time series modelling procedures try to forecast future values from lagged ones. Thus, the selection of the relevant lagged values to be used is a key step. In this paper, a new consensus method for the selection of relevant lagged values of a time series is introduced: feature ranking aggregated selection (FRASel). The main contribution of this feature selection method is the definition of a consensus decision making mechanism based on aggregation and expressed as a simple rule. In FRASel, the selected subset of lagged values is decided by the application of an aggregation criterion to the results of different flavours of feature ranking methods, applied from different approaches. A thorough empirical analysis is carried out to assess the performance of FRASel. The statistical significance of the experimental results is also analysed through the application of non-parametric statistical tests.     

类别增量学习研究进展和性能评价

机器学习技术成功地应用于计算机视觉、自然语言处理和语音识别等众多领域.然而,现有的大多数机器学习模型在部署后类别和参数是固定的,只能泛化到训练集中出现的类别,无法增量式地学习新类别.在实际应用中,新的类别或任务会源源不断地出现,这要求模型能够像人类一样在较好地保持已有类别知识的基础上持续地学习新类别知识.近年来新兴的类别增量学习研究方向,旨在使得模型能够在开放、动态的环境中持续学习新类别的同时保持对旧类别的判别能力(防止“灾难性遗忘”).本文对类别增量学习(Class-incremental learning, CIL)方法进行了详细综述.根据克服遗忘的技术思路,将现有方法分为基于参数正则化、基于知识蒸馏、基于数据回放、基于特征回放和基于网络结构的五类方法,对每类方法的优缺点进行了总结.此外,本文在常用数据集上对代表性方法进行了实验评估,并通过实验结果对现有算法的性能进行了比较分析.最后,对类别增量学习的研究趋势进行展望.     

A review and analysis of “graph theoretical-matrix permanent” approach to decision making with example applications

In this paper, a multiple attribute decision making technique is explained in detail and its existing applications are reviewed/analyzed first time in the literature. The technique was originated from combinatorial mathematics and it is based on the graph theory and matrix algebra and has some desirable properties like “ability to model criteria interactions”, “ability to generate hierarchical models etc.” for modeling and solving complex decision making problems. In order to enable a better understanding of the technique, two illustrative examples (new graduates’ industry sector preferences and supermarket location selection) with crisp and fuzzy values are also modeled and solved in the present study.     

User demographics prediction based on mobile data

Demographics prediction is an important component of user profile modeling. The accurate prediction of users’ demographics can help promote many applications, ranging from web search, personalization to behavior targeting. In this paper, we focus on how to predict users’ demographics, including “gender”, “job type”, “marital status”, “age” and “number of family members”, based on mobile data, such as users’ usage logs, physical activities and environmental contexts. The core idea is to build a supervised learning framework, where each user is represented as a feature vector and users’ demographics are considered as prediction targets. The most important component is to construct features from raw data and then supervised learning models can be applied. We propose a feature construction framework, CFC (contextual feature construction), where each feature is defined as the conditional probability of one user activity under the given contexts. Consequently, besides employing standard supervised learning models, we propose a regularized multi-task learning framework to model different kinds of demographics predictions collectively. We also propose a cost-sensitive classification framework for regression tasks, in order to benefit from the existing dimension reduction methods. Finally, due to the limited training instances, we employ ensemble to avoid overfitting. The experimental results show that the framework achieves classification accuracies on “gender”, “job” and “marital status” as high as 96%, 83% and 86%, respectively, and achieves Root Mean Square Error (RMSE) on “age” and “number of family members” as low as 0.69 and 0.66 respectively, under the leave-one-out evaluation.     

Issues of numerical accuracy and stability in inverse simulation

Inverse simulation algorithms based on integration have been widely applied to predict the control input time histories required for aircraft to follow ideally defined manoeuvres. Several different inverse simulation algorithms are available but these different methods are all subject to a number of numerical and stability problems, such as high frequency oscillation effects and also lower frequency oscillatory phenomena termed “constraint oscillations”. Difficulties can also arise in applications involving discontinuous manoeuvres, discontinuities within the model or input constraints involving actuator saturation. This paper has shown that the dynamic response properties of the internal system are the cause of the so-called “constraint oscillation” phenomenon. In addition, a new inverse simulation approach based on the constrained derivative-free Nelder–Mead search-based optimisation method has been developed to eliminate problems of discontinuities and saturation. Simulation studies involving nonlinear ship models suggest that this new approach leads to improved properties in terms of convergence and numerical stability.     

An extended model for remaining time prediction in manufacturing systems using process mining

The ability to predict the remaining cycle-time in industrial environments is of major concern among production managers. An accurate prediction would enable managers to handle undesired situations with more control, thereby preventing future losses. However, making such predictions is no trivial task: there are many methods available to cope with this problem, including a recent research stream in process mining. Process mining provides tools for automated discovery of process models from event logs, and eventually, extend those models in driving predictions. In general, predictive models in process mining generally deals with business processes, and not directly with the industrial environment, which contains a full prism of particularities. In this paper we propose a hybrid predictive model based on transition-systems and statistical regression which is “product-oriented”, tailored to better predict online cycle-times on industrial environments. We propose a weight for each method, optimized by a linear programming model. We tested our new approach on an artificially created log that emulates an industrial environment, and on a real manufacture log. Results showed that our approach provides better accuracy measures for both test instances.     

The role of proteomics in understanding biological mechanisms of sepsis

Sepsis is a systemic inflammatory state caused by infection. Complications of this infection with multiple organ failure lead to more lethal conditions, such as severe sepsis and septic shock. Sepsis is one of the leading causes of US deaths. Novel biomarkers with high sensitivity and specificity may be helpful for early diagnosis of sepsis and for improvement of patient outcomes through the development of new therapies. Mass spectrometry-based proteomics offers powerful tools to identify such biomarkers and furthermore to give insight to fundamental mechanisms of this clinical condition. In this review, we summarize findings from proteomics studies of sepsis and how their applications have provided more understanding into the pathogenesis of septic infection. Literatures related to “proteomics”, “sepsis”, “systemic inflammatory response syndrome”, “severe sepsis”, “septic infection”, and “multiple organ dysfunction syndrome” were searched using PubMed. Findings about neonatal and adult sepsis are discussed separately. Within the adult sepsis studies, results are grouped based on the models (e.g., human or animal). Across investigations in clinical populations and in rodent and mammalian animal models, biological pathways, such as inflammatory and acute phase response, coagulation, complement, mitochondrial energy metabolism, chaperones, and oxidative stress, are altered at the protein level. These proteomics studies have discovered many novel biomarker candidates of septic infection. Validation the clinical use of these biomarker candidates may significantly impact the diagnosis and prognosis of sepsis. In addition, the molecular mechanisms revealed by these studies may also guide the development of more effective treatments.     

Semantics enactment for interoperability assessment in enterprise information systems

The grown complexity of the modern enterprise poses a series of challenges, among them keeping competitiveness in the fast changing environment in which the enterprise evolves. Addressing enterprise integration is considered as a key to achieve the goal of any enterprise either it is a single or a networked enterprise. Enterprise modelling is a prerequisite to enable the common understanding of the enterprises and its various interactions in order to “provide the right information, at the right time, at the right place”. However, problems often emerge from a lack of understanding of the semantics of the elaborated models resulting from various modelling experience based on different methods and tools. This paper describes the challenges associated to semantics enactment in information systems models. To facilitate this enactment, it proposes an approach based on a fact-oriented modelling perspective. Then, it also provides an algorithm to automatically build semantic aggregates that help in highlighting enterprise models core embedded semantics. A case study on the field of B2M interoperability is performed in order to illustrate the application of the presented approach.     

“互联网+”背景下移动应用开发类课程一体化教学改革与实践

“互联网+”新经济形态下,分析移动应用开发行业发展趋势及课程教学现状,形成“标准引领、分层递进”的教学内容,提出“双方参与、全程多维、产教融合”的合作课程育人模式,创新“情境化项目教学为主,移动教学为辅”的课程教学方法,搭建融入“数字化校园+云计算+大数据”的课程移动学习平台,建立基于大数据技术的课堂教学行为评价标准,实施融创新创业教育的课程教学实践,通过一体化的教学改革,提高了课程教学的时效性,增强了就业能力。     

Reduced Modeling of Unknown Trajectories

This paper deals with model order reduction of parametrical dynamical systems. We consider the specific setup where the distribution of the system’s trajectories is unknown but the following two sources of information are available: (i) some “rough” prior knowledge on the system’s realisations; (ii) a set of “incomplete” observations of the system’s trajectories. We propose a Bayesian methodological framework to build reduced-order models (ROMs) by exploiting these two sources of information. We emphasise that complementing the prior knowledge with the collected data provably enhances the knowledge of the distribution of the system’s trajectories. We then propose an implementation of the proposed methodology based on Monte-Carlo methods. In this context, we show that standard ROM learning techniques, such e.g., proper orthogonal decomposition or dynamic mode decomposition, can be revisited and recast within the probabilistic framework considered in this paper. We illustrate the performance of the proposed approach by numerical results obtained for a standard geophysical model.     

Function evaluation by incremental computation,with applications

Simulation of physical systems often requires repetitive evaluation of functions such as sine, cosine, exponential etc. The arguments of these functions are physical quantities, which usually change very little from one computation cycle to the next. An approach to function evaluation is proposed, which utilizes the “slowness” property in order to reduce computation time. This approach — “incremental process” — is, in a sense, a numerical solution of a differential equation whose solution is the desired function. The main drawback of incremental methods lies in the possibility of error propagation and accumulation. This phenomenon is very noticeable when the argument oscillates around a fixed value, since the errors grow while the true solution is nearly constant (“rectification” error). It was proposed that “reversible” incremental processes may exist, which will limit error propagation, in certain situations, by regaining their (exact) initial value whenever their argument returns to its initial value. We show that such reversible processes cannot exist for transcendental functions if their argument increments may assume any value within a permissible range. Placing certain reasonable restrictions on the increment values does lead, however, to algorithms which save computer time in comparison with conventional function evaluation algorithms. Several examples are presented.     

Multi-fidelity POD surrogate-assisted optimization: Concept and aero-design study

We combine multiple sets of variable-precision full-field simulations within a single surrogate model. The approach is based on an original formulation of the “Constrained Proper Orthogonal Decomposition” (C-POD), interpolating precise, albeit costly, high-fidelity data and approximating abondant, yet less accurate, lower-fidelity data. We compute the optimal high-dimensional subspace spanning the sparse high-fidelity full-field solutions and refine the output subspace definition thanks to the orthogonal information contained in abondant low-fidelity full-field solutions. We then build “hierarchised” multi-fidelity surrogate models based on the previously refined subspace and giving a fast estimation of the high-fidelity full-field solution of any new location in the design space. The proposed model is illustrated by exploring the prediction of an analytical 2D functional space on the one hand, and demonstrated by studying the efficiency of a 1.5-stage low-pressure compressor on the other.     

Ten years of visualization of business process models: A systematic literature review

Business process modeling is an essential task in business process management. Process models that are comprehensively understood by business stakeholders allow organizations to profit from this field. In this work, we report what is being investigated in the topic “visualization of business process models”, since visualization is known as improving perception and comprehension of structures and patterns in datasets. We performed a systematic literature review through which we selected and analyzed 46 papers from two points of view. Firstly, we observed the similarities between the papers regarding their main scope. From this observation we classified the papers into six categories: “Augmentation of existing elements”, “Creation of new elements”, “Exploration of the 3D space”, “Information visualization”, “Visual feedback concerning problems detected in process models” and “Perspectives”. The less explored categories and which could represent research challenges for further exploration are “Visual feedback” and “Information visualization”. Secondly, we analyzed the papers based on a well-known visualization analysis framework, which allowed us to obtain a high-level point of view of the proposals presented in the literature and could identify that few authors explore user interaction features in their works. Besides that, we also found that exactly half of the papers base their proposals on BPMN and present results from evaluation or validation. Since BPMN is an ISO standard and there are many tools based on BPMN, there should be more research intending to improve the knowledge around this topic. We expect that our results inspire researchers for further work aiming at bringing forward the field of business process model visualization, to have the advantages of information visualization helping the tasks of business process modeling and management.     

Modelling domestic water demand: An agent based approach

The urban water system is a complex adaptive system consisting of technical, environmental and social components which interact with each other through time. As such, its investigation requires tools able to model the complete socio-technical system, complementing “infrastructure-centred” approaches. This paper presents a methodology for integrating two modelling tools, a social simulation model and an urban water management tool. An agent based model, the Urban Water Agents' Behaviour, is developed to simulate the domestic water users’ behaviour in response to water demand management measures and is then coupled to the Urban Water Optioneering Tool to calculate the evolution of domestic water demand by simulating the use of water appliances. The proposed methodology is tested using, as a case study, a major period of drought in Athens, Greece. Results suggest that the coupling of the two models provides new functionality for water demand management scenarios assessment by water regulators and companies.