Integrating planning and learning: the PRODIGY architecture

Abstract

Planning is a complex reasoning task that is well suited for the study of improving performance and knowledge by learning, i.e. by accumulation and interpretation of planning experience. PRODIGY is an architecture that integrates planning with multiple learning mechanisms. Learning occurs at the planner's decision points and integration in PRODIGY is achieved via mutually interpretable knowledge structures. This article describes the PRODIGY planner, briefly reports on several learning modules developed earlier along the project, and presents in more detail two recently explored methods to learn to generate plans of better quality. We introduce the techniques, illustrate them with comprehensive examples, and show preliminary empirical results. The article also includes a retrospective discussion of the characteristics of the overall PRODIGY architecture and discusses their evolution within the goal of the project of building a large and robust integrated planning and learning system.     

Cima: An interactive concept learning system for end-user applications

Personalizable software agents will learn new tasks from their users. In many cases the most appropriate way for users to teach is to demonstrate examples. Learning complex concepts from examples alone is hard, but agents can exploit other forms of instruction that users might give, ranging from yes/no responses to ambiguous, incomplete hints. Agents can also exploit background knowledge customized for applications such as drawing, word processing, and form filling. The Cima system learns generalized rules for classifying, generating, and modifying data, given examples, hints, and background knowledge. It copes with the ambiguity of user instructions by combining evidence from these sources. A dynamic bias manager generates candidate features (attribute values, functions, or relations) from which the learning algorithm selects relevant ones and forms appropriate rules. When tested on dialogs observed in a prior user study on a simulated interface agent, the system achieved 95% of the learning efficiency observed in that study.     

Case-based analysis in user requirements modelling for knowledge construction

ContextLearning can be regarded as knowledge construction in which prior knowledge and experience serve as basis for the learners to expand their knowledge base. Such a process of knowledge construction has to take place continuously in order to enhance the learners’ competence in a competitive working environment. As the information consumers, the individual users demand personalised information provision which meets their own specific purposes, goals, and expectations.ObjectivesThe current methods in requirements engineering are capable of modelling the common user’s behaviour in the domain of knowledge construction. The users’ requirements can be represented as a case in the defined structure which can be reasoned to enable the requirements analysis. Such analysis needs to be enhanced so that personalised information provision can be tackled and modelled. However, there is a lack of suitable modelling methods to achieve this end. This paper presents a new ontological method for capturing individual user’s requirements and transforming the requirements onto personalised information provision specifications. Hence the right information can be provided to the right user for the right purpose.MethodAn experiment was conducted based on the qualitative method. A medium size of group of users participated to validate the method and its techniques, i.e. articulates, maps, configures, and learning content. The results were used as the feedback for the improvement.ResultThe research work has produced an ontology model with a set of techniques which support the functions for profiling user’s requirements, reasoning requirements patterns, generating workflow from norms, and formulating information provision specifications.ConclusionThe current requirements engineering approaches provide the methodical capability for developing solutions. Our research outcome, i.e. the ontology model with the techniques, can further enhance the RE approaches for modelling the individual user’s needs and discovering the user’s requirements.     

A Framework for the Generation of Adaptive Courses Based on Semantic Metadata

This approach proposes the creation and management of adaptive learning systems by combining component technology, semantic metadata, and adaptation rules. A component model allows interaction among components that share consistent assumptions about what each provides and each requires of the other. It allows indexing, using, reusing, and coupling of components in different contexts powering adaptation. Our claim is that semantic metadata are required to allow a real reusing and assembling of educational component. Finally, a rule language is used to define strategies to rewrite user query and user model. The former allows searching components developing concepts not appearing in the user query but related with user goals, whereas the last allow inferring user knowledge that is not explicit in user model.John Freddy Duitama received his M.Sc. degree in system engineering from the University of Antioquia -Colombia (South America). He is currently a doctoral candidate in the GET – Institut National des Télécommunications, Evry France. This work is sponsored by the University of Antioquia, where he is assistant professor.His research interest includes semantic web and web-based learning systems, educational metadata and learning objects.Bruno Defude received his Ph.D. in Computer Science from the University of Grenoble (I.N.P.G) in 1986. He is currently Professor in the Department of Computer Science at the GET - Institut National des Télécommunications, Evry France where he leads the SIMBAD project (Semantic Interoperability for MoBile and ADaptive applications).His major field of research interest is databases and semantic web, specifically personalized data access, adaptive systems, metadata, interoperability and semantic Peer-to-peer systems with elearning as a privileged application area.He is a member of ACM SIGMOD.Amel Bouzeghoub received a degree of Ph.D. in Computer Sciences at Pierre et Marie Curie University, France.In 2000, she joined the Computer Sciences Department of GET-INT (Institut National des Telecommunications) at Evry (France) as an associate professor.Her research interests include topics related to Web-based Learning Systems, Semantic Metadata for learning resources, Adaptive Learning Systems and Intelligent Tutoring Systems.Claire Lecocq received an Engineer Degree and a Ph.D. in Computer Sciences respectively in 1994 and 1999. In 1997, she joined the Computer Sciences Department at GET-INT (Institut National des Télécommunications) of Evry, France, as an associate professor. Her first research interests included spatial databases and visual query languages. She is now working on adaptive learning systems, particularly on semantic metadata and user models.     

A prototype-based method for classification with time constraints: a case study on automated planning

The main goal of Nearest Prototype Classification is to reduce storage space and retrieval time of classical Instance-Based Learning (IBL) algorithms. This motivation is higher in relational data since relational distance metrics are much more expensive to compute than classical distances like Euclidean distance. In this paper, we present an algorithm to build Relational Nearest Prototype Classifiers (RNPCs). When compared with Relational Instance-Based Learning (Relational IBL or RIBL) approaches, the algorithm is able to dramatically reduce the number of instances by selecting the most relevant prototypes, maintaining similar accuracy. The number of prototypes is obtained automatically by the algorithm, although it can also be bound by the user. In this work, we also show an application of RNPC for automated planning. Specifically, we describe a modeling task where a relational policy is built following an IBL approach. This approach uses the decisions taken by a planning system as learning examples. We show that when the number of learning examples is reduced with RNPC, the resulting policy is able to scale up better than the original planning system.     

Animated interface agent applying atms-based multimodal input interpretation

Two requirements should be met in order to develop a practical multimodal interface system , i . e ., ( 1 ) integration of delayed arrival of data and ( 2 ) elimination of ambiguity in recognition results of each modality . This paper presents an efficient and generic methodology for interpretation of multimodal input to satisfy these requirements . The proposed methodology can integrate delayed - arrival data satisfactorily and efficiently interpret multimodal input that contains ambiguity . In the input interpretation the multimodal interpretation process is regarded as hypothetical reasoning , and the control mechanismof interpretation is formalized by applying the assumption - based truth maintenance system ( ATMS ). The proposed method is applied to an interface agent system that accepts multimodal input consisting of voice and direct indication gesture on a touch display . The systemcommunicates to the user through a human - like interface agent's three - dimensional motion image with facial expressions , gestures , and a synthesized voice .     

An Authentication and Authorization Solution for a Multiplatform Cloud Environment

ABSTRACT

Providing security at all levels within the multiplatform cloud-computing environment is has not been properly solved due to a variety of problems arising from technical and human-based sources. This paper presents an authentication-and-authorization solution based on the Single Sign-On (SSO) approach for cloud-service users and administrators in a multiplatform environment. The system developed enables user authentication for clouds provided as Infrastructure as a Service system built up from different OS systems. The solution enables the use of different services based on credentials that are authenticated only once and enable simple and efficient administration of the relevant data. The paper briefly presents the problem of user authentication in cloud services from the security aspect and defines the user and system administrator requirements for a secure and efficient authentication system. The implemented solution for two different platforms and the associated OS, one proprietary (WMware) and one open-source (OpenStack), is briefly described.     

PVA: A Self-Adaptive Personal View Agent

In this paper, we present PVA, an adaptive personal view information agent system for tracking, learning and managing user interests in Internet documents. PVA consists of three parts: a proxy, personal view constructor, and personal view maintainer. The proxy logs the user's activities and extracts the user's interests without user intervention. The personal view constructor mines user interests and maps them to a class hierarchy (i.e., personal view). The personal view maintainer synchronizes user interests and the personal view periodically. When user interests change, in PVA, not only the contents, but also the structure of the user profile are modified to adapt to the changes. In addition, PVA considers the aging problem of user interests. The experimental results show that modulating the structure of the user profile increases the accuracy of a personalization system.     

结合因子图的多目的地地图布局优化

目的提出一种结合因子图的多目的地地图生成方法。方法首先,由用户选择多个感兴趣的目的地,系统根据相应规则自动地选择与目的地最相关的路线。然后,通过定义一组衡量布局质量的约束规则,采用因子图方法将定义的每条规则编码成因子,并采用Metropolis Hastings算法对由因子图构建得到的目标分布函数进行采样得到符合约束规则的多目的地地图。结果实验结果表明,使用这种方法得到的多目的地地图,可以在同一显示空间中显示多个目的地之间的道路信息,同时又保留了各目的地区域之间的拓扑和空间关系。结论提出的多目的地地图能有效地为用户提供导航,解决了当前在线地图无法在同一视野中为用户提供空间距离较远的区域道路信息的问题。     

A vision system for symbolic interpretation of dynamic scenes using arsom

We describe an artificial high-level vision system for the symbolic interpretation of data coming from a video camera that acquires the image sequences of moving scenes. The system is based on ARSOM neural networks that learn to generate the perception-grounded predicates obtained by image sequences. The ARSOM neural networks also provide a three-dimensional estimation of the movements of the relevant objects in the scene. The vision system has been employed in two scenarios: the monitoring of a robotic arm suitable for space operations, and the surveillance of an electronic data processing (EDP) center.     

User profiling for the melvil knowledge retrieval system

Melvil is an ontology-based knowledge retrieval platform that provides a three-dimensional visualization of search results. The user can tailor the presentation of the search results to his or her preferences by changing the settings of various parameters on the screen. In this paper, we report on a prototype implementation of a user profiling device that learns to predict appropriate settings for these parameters for the current search results based on previous experiences. In a preliminary study, we evaluated several off-the-shelf machine learning algorithms on parts of the problem. The final implementation required the flexibility of handling both regression and classification problems, being able to deal with set-valued input and output attributes, as well as incorporating Melvil 's ontologies for the respective application domain. Thus, we selected a nearest-neighbor approach for the prototype implementation. An evaluation on off-line data collected from several users showed a satisfactory performance.     

Incremental planning using conceptual graphs

Abstract

When performing a planning or design task in many domains it is often difficult to specify in advance what the precise goals are. It is therefore useful to have a system in which the planning process is performed interactively, with the solution approaching the users' intent incrementally through iterations of the planning process. A planning system intended to function in this way must be able to take goal specifications interactively rather than all at once at the beginning of the planning process. The planning process then becomes one of satisfying new goals as they are given by the user, modifying as little as possible the results of previous planning work. Incremental planning is an approach to interactive planning problems that allows a system to create a plan incrementally, modifying a previous plan to satisfy new or more precise goal specifications. In this paper we present an incremental planning system called the general constraint system (GCS) that is based on the conceptual programming environment (CP) developed at New Mexico State University and we show an example of the use of the system for a simple civil engineering design problem     

Use and non‐use of VDTs—organization of terminal work: Research findings from Swedish cross‐site studies in the field of office automation

A group of users in Copenhagen were asked to evaluate how important a number of user interface characteristics were for them. The results show high importance of efficient daily use and of possibilities for exploratory learning while tutorial materials were of less importance. Users also were asked to evaluate four usability aspects of a number of popular programs. Results show that the quality pleasant to work with has the largest impact on evaluations of overall user‐friendliness while users seem able to view usability independently from the number of features in an application.     

Incremental multistrategy learning for document processing

This work presents the application of a multistrategy approach to some document processing tasks. The application is implemented in an enhanced version of the incremental learning system INTHELEX. This learning module has been embedded as a learning component in the system architecture of the EU project COLLATE, which deals with the annotation of cultural heritage documents. Indeed, the complex shape of the material handled in the project has suggested that the addition of multistrategy capabilities is needed to improve effectiveness and efficiency of the learning process. Results proving the benefits of these strategies in specific classfication tasks are reported in the experimentation presented in this work.     

A planning aid for human factors evaluation practice

Abstract

The work reported here attempts to address Human-Computer Interaction (HCI) design problems by the creation of support for the conceptualization of such problems during evaluation. This support takes the form of a planning aid intended to aid novice human factors practitioners(recently qualified graduates, for example) to evaluate interactive work systems. The planning aidprovides a structure for relating and recruiting techniques used in Human Factors (HF) evaluations. It incorporates relevant information for planning an evaluation (e.g., evaluation methods themselves), and offers advice in the form of heuristics about the use of the methods, their selection, and configuration. The output of the planning aid is an evaluation plan.This paper reports the development of the planning aid, and illustrates its application with a case study. Two assessments of the planning aid with novice HF practitioners are also presented and discussed.     

HCI and business practices in a collaborative method for augmented reality systems

ContextEvery interactive system is composed of a functional core and a user interface. However, the software engineering (SE) and human–computer interaction (HCI) communities do not share the same methods, models or tools. This usually induces a large work overhead when specialists from the two domains try to connect their applicative studies, especially when developing augmented reality systems that feature complex interaction cores.ObjectiveWe present in this paper the essential activities and concepts of a development method integrating the SE and HCI development practices, from the specifications down to the design, as well as their application on a case study.MethodThe efficiency of the method was tested in a qualitative study involving four pairs of SE and HCI experts in the design of an application for which an augmented reality interaction would provide better user performance than a classic interactive system. The effectivity of the method was evaluated in a qualitative study comparing the quality of three implementations of the same application fragment (based on the same analysis model), using software engineering metrics.ResultsThe first evaluation confirmed the ease of use of our method and the relevance of our tools for guiding the design process, but raised concerns on the handling of conflicting collaborative activities. The second evaluation gave indications that the structure of the analysis model facilitates the implementation of quality software (in terms of coupling, stability and complexity).ConclusionIt is concluded that our method enables design teams with different backgrounds in application development to collaborate for integrating augmented reality applications with information systems. Areas of improvement are also described.     

Machine learning: A tool to support usability?

One problem facing designers of interactive systems is catering to the wide range of users who will use a particular application. Understanding the user is critical to designing a usable interface. There are a number of ways of addressing this problem, including improved design methodologies using ''intuitive'' interface styles, adaptive interfaces, and better training and user support materials. In this article, we argue that each of these solutions involves pattern recognition in one form or another and that machine learning can therefore aid designers of interactive systems in these areas. We report on experiments that demonstrate the potential of machine learning to user modeling that has application to two of these areas in particular: adaptive systems and design methodologies.     

Inductive program synthesis for therapy plan generation

Planning is investigated in an area where classical STRIPS-like approaches usually fail. The application domain is therapy (i.e. repair) for complex dynamic processes. The peculiarities of this domain are discussed in some detail for convincingly developing the characteristics of the inductive planning approach presented. Plans are intended to be run for process therapy. Thus, plans are programs. Because of the unavoidable vagueness and uncertainty of information about complex dynamic processes in the case of disturbance, therapy plan generation turns out to be inductive program synthesis. There is developed a graph-theoretically based approach to inductive therapy plan generation. This approach is investigated from the inductive inference perspective. Particular emphasis is put on consistent and incremental learning of therapy plans. Basic application scenarios are developed and compared to each other. The inductive inference approach is invoked to develop and investigate a couple of planning algorithms. The core versions of these algorithms are successfully implemented in Lisp and Prolog. The work has been partially supported by the German Federal Ministry for Research and Technology (BMFT) within the Joint Project (BMFT-Verbundprojekt)Wiscon onDevelopment of Methods for Intelligent Monitoring and Control under contract no. 413-4001-01 IW 204 B. Additionally, the second author’s work in learning theory received some support from the German Federal Ministry for Research and Technology (BMFT) within the Joint Project (BMFT-Verbundprojekt)Gosler onAlgorithmic Learning for Knowledge-Based Systems under contract no. 413-4001-01 IW 101 A. Oksana Arnold: She graduated from Leipzig University of Technology in 1990 with a Master’s Thesis on a rule interpreter for default reasoning. She received her PhD. in Computer Science in 1996 on therapy control for complex dynamic processes within a knowledge-based process supervision and control system. Recently, She works at the University of Leipzig within a research project on information and communication technologies for virtual enterprises. Her main scientific interest is both in knowledge-based process supervision and control, where she did a pioneering work on therapy plan generation, and in flexible information systems for new generation business applications. Klaus P. Jantke: He graduated from Humboldt University Berlin with a Master’s Thesis in 1975. He received his Ph. D. in Computer Science in 1979 and his Habilitation at Humboldt in 1984. He worked as the Head of a Research Laboratory in Theoretical Computer Science and as a Vice-Director of the Computing Center at Humboldt University. Since 1987, Dr. Jantke is full professor at Leipzig University of Technology. His main research interest is in algorithmic learning theory. Besides this, he contributes to case-based reasoning, where his special interest is in learning issues and in structural similarity, and to knowledge-based process supervision and control, especially to planning. Dr. Jantke is member of the ACM, the EATCS, and the GI.     

A Control Architecture for Robot Swarms (AMEB)

Abstract

A control architecture for heterogeneous robot swarms (AMEB) is proposed in this work. The architecture manages the processes that occur in the system and its main objectives are to allow the emergence and self-organization in the group. It is structured in three levels: an individual level, a collective level under the philosophy of emergent behavior, and a level for the management of learning and knowledge. The architecture includes a behavioral component that allows the inclusion of emotions in the members of the swarm. Finally, it described a method for verifying the occurrence of the emergence in the swarm, using fuzzy cognitive maps.