Network-based stochastic semisupervised learning

Semisupervised learning is a machine learning approach that is able to employ both labeled and unlabeled samples in the training process. In this paper, we propose a semisupervised data classification model based on a combined random-preferential walk of particles in a network (graph) constructed from the input dataset. The particles of the same class cooperate among themselves, while the particles of different classes compete with each other to propagate class labels to the whole network. A rigorous model definition is provided via a nonlinear stochastic dynamical system and a mathematical analysis of its behavior is carried out. A numerical validation presented in this paper confirms the theoretical predictions. An interesting feature brought by the competitive-cooperative mechanism is that the proposed model can achieve good classification rates while exhibiting low computational complexity order in comparison to other network-based semisupervised algorithms. Computer simulations conducted on synthetic and real-world datasets reveal the effectiveness of the model.

New Techniques in Deep Representation Learning

Thesis (Ph.D.)--University of Washington, 2016-06

Using the EM Algorithm to Train Neural Networks: Misconceptions and a New Algorithm for Multiclass Classification

The expectation-maximization (EM) algorithm has been of considerable interest in recent years as the basis for various algorithms in application areas of neural networks such as pattern recognition. However, there exists some misconceptions concerning its application to neural networks. In this paper, we clarify these misconceptions and consider how the EM algorithm can be adopted to train multilayer perceptron (MLP) and mixture of experts (ME) networks in applications to multiclass classification. We identify some situations where the application of the EM algorithm to train MLP networks may be of limited value and discuss some ways of handling the difficulties. For ME networks, it is reported in the literature that networks trained by the EM algorithm using iteratively reweighted least squares (IRLS) algorithm in the inner loop of the M-step, often performed poorly in multiclass classification. However, we found that the convergence of the IRLS algorithm is stable and that the log likelihood is monotonic increasing when a learning rate smaller than one is adopted. Also, we propose the use of an expectation-conditional maximization (ECM) algorithm to train ME networks. Its performance is demonstrated to be superior to the IRLS algorithm on some simulated and real data sets.

A Framework of Fuzzy Diagnosis

Fault diagnosis has become an important component in intelligent systems, such as intelligent control systems and intelligent eLearning systems. Reiter's diagnosis theory, described by first-order sentences, has been attracting much attention in this field. However, descriptions and observations of most real-world situations are related to fuzziness because of the incompleteness and the uncertainty of knowledge, e. g., the fault diagnosis of student behaviors in the eLearning processes. In this paper, an extension of Reiter's consistency-based diagnosis methodology, Fuzzy Diagnosis, has been proposed, which is able to deal with incomplete or fuzzy knowledge. A number of important properties of the Fuzzy diagnoses schemes have also been established. The computing of fuzzy diagnoses is mapped to solving a system of inequalities. Some special cases, abstracted from real-world situations, have been discussed. In particular, the fuzzy diagnosis problem, in which fuzzy observations are represented by clause-style fuzzy theories, has been presented and its solving method has also been given. A student fault diagnostic problem abstracted from a simplified real-world eLearning case is described to demonstrate the application of our diagnostic framework.

Facilitating cross-organisational workflows with a workflow view approach

Interconnecting business processes across systems and organisations is considered to provide significant benefits, such as greater process transparency, higher degrees of integration, facilitation of communication, and consequently higher throughput in a given time interval. However, to achieve these benefits requires tackling constraints. In the context of this paper these are privacy-requirements of the involved workflows and their mutual dependencies. Workflow views are a promising conceptional approach to address the issue of privacy; however this approach requires addressing the issue of interdependencies between workflow view and adjacent private workflow. In this paper we focus on three aspects concerning the support for execution of cross-organisational workflows that have been modelled with a workflow view approach: (i) communication between the entities of a view-based workflow model, (ii) their impact on an extended workflow engine, and (iii) the design of a cross-organisational workflow architecture (CWA). We consider communication aspects in terms of state dependencies and control flow dependencies. We propose to tightly couple private workflow and workflow view with state dependencies, whilst to loosely couple workflow views with control flow dependencies. We introduce a Petri-Net-based state transition approach that binds states of private workflow tasks to their adjacent workflow view-task. On the basis of these communication aspects we develop a CWA for view-based cross-organisational workflow execution. Its concepts are valid for mediated and unmediated interactions and express no choice of a particular technology. The concepts are demonstrated by a scenario, run by two extended workflow management systems. (C) 2004 Elsevier B.V. All rights reserved.

An expert system on design of liquid-retaining structures with blackboard architecture

The design of liquid-retaining structures involves many decisions to be made by the designer based on rules of thumb, heuristics, judgement, codes of practice and previous experience. Structural design problems are often ill structured and there is a need to develop programming environments that can incorporate engineering judgement along with algorithmic tools. Recent developments in artificial intelligence have made it possible to develop an expert system that can provide expert advice to the user in the selection of design criteria and design parameters. This paper introduces the development of an expert system in the design of liquid-retaining structures using blackboard architecture. An expert system shell, Visual Rule Studio, is employed to facilitate the development of this prototype system. It is a coupled system combining symbolic processing with traditional numerical processing. The expert system developed is based on British Standards Code of Practice BS8007. Explanations are made to assist inexperienced designers or civil engineering students to learn how to design liquid-retaining structures effectively and sustainably in their design practices. The use of this expert system in disseminating heuristic knowledge and experience to practitioners and engineering students is discussed.

Ontological evaluation of enterprise systems interoperability using ebXML

Enterprise systems interoperability (ESI) is an important topic for business currently. This situation is evidenced, at least in part, by the number and extent of potential candidate protocols for such process interoperation, viz., ebXML, BPML, BPEL, and WSCI. Wide-ranging support for each of these candidate standards already exists. However, despite broad acceptance, a sound theoretical evaluation of these approaches has not yet been provided. We use the Bunge-Wand-Weber (BWW) models, in particular, the representation model, to provide the basis for such a theoretical evaluation. We, and other researchers, have shown the usefulness of the representation model for analyzing, evaluating, and engineering techniques in the areas of traditional and structured systems analysis, object-oriented modeling, and process modeling. In this work, we address the question, what are the potential semantic weaknesses of using ebXML alone for process interoperation between enterprise systems? We find that users will lack important implementation information because of representational deficiencies; due to ontological redundancy, the complexity of the specification is unnecessarily increased; and, users of the specification will have to bring in extra-model knowledge to understand constructs in the specification due to instances of ontological excess.

How do practitioners use conceptual modeling in practice?

Much research has been devoted over the years to investigating and advancing the techniques and tools used by analysts when they model. As opposed to what academics, software providers and their resellers promote as should be happening, the aim of this research was to determine whether practitioners still embraced conceptual modeling seriously. In addition, what are the most popular techniques and tools used for conceptual modeling? What are the major purposes for which conceptual modeling is used? The study found that the top six most frequently used modeling techniques and methods were ER diagramming, data flow diagramming, systems flowcharting, workflow modeling, UML, and structured charts. Modeling technique use was found to decrease significantly from smaller to medium-sized organizations, but then to increase significantly in larger organizations (proxying for large, complex projects). Technique use was also found to significantly follow an inverted U-shaped curve, contrary to some prior explanations. Additionally, an important contribution of this study was the identification of the factors that uniquely influence the decision of analysts to continue to use modeling, viz., communication (using diagrams) to/from stakeholders, internal knowledge (lack of) of techniques, user expectations management, understanding models' integration into the business, and tool/software deficiencies. The highest ranked purposes for which modeling was undertaken were database design and management, business process documentation, business process improvement, and software development. (c) 2005 Elsevier B.V. All rights reserved.

Stability criteria for unsupervised temporal association networks

A biologically realizable, unsupervised learning rule is described for the online extraction of object features, suitable for solving a range of object recognition tasks. Alterations to the basic learning rule are proposed which allow the rule to better suit the parameters of a given input space. One negative consequence of such modifications is the potential for learning instability. The criteria for such instability are modeled using digital filtering techniques and predicted regions of stability and instability tested. The result is a family of learning rules which can be tailored to the specific environment, improving both convergence times and accuracy over the standard learning rule, while simultaneously insuring learning stability.

Selecting optimal instantiations of data models - Theory and validation of an ex ante approach

The schema of an information system can significantly impact the ability of end users to efficiently and effectively retrieve the information they need. Obtaining quickly the appropriate data increases the likelihood that an organization will make good decisions and respond adeptly to challenges. This research presents and validates a methodology for evaluating, ex ante, the relative desirability of alternative instantiations of a model of data. In contrast to prior research, each instantiation is based on a different formal theory. This research theorizes that the instantiation that yields the lowest weighted average query complexity for a representative sample of information requests is the most desirable instantiation for end-user queries. The theory was validated by an experiment that compared end-user performance using an instantiation of a data structure based on the relational model of data with performance using the corresponding instantiation of the data structure based on the object-relational model of data. Complexity was measured using three different Halstead metrics: program length, difficulty, and effort. For a representative sample of queries, the average complexity using each instantiation was calculated. As theorized, end users querying the instantiation with the lower average complexity made fewer semantic errors, i.e., were more effective at composing queries. (c) 2005 Elsevier B.V. All rights reserved.

Intelligent agent supported personalization for virtual learning environments

Virtual learning environments (VLEs) are computer-based online learning environments, which provide opportunities for online learners to learn at the time and location of their choosing, whilst allowing interactions and encounters with other online learners, as well as affording access to a wide range of resources. They have the capability of reaching learners in remote areas around the country or across country boundaries at very low cost. Personalized VLEs are those VLEs that provide a set of personalization functionalities, such as personalizing learning plans, learning materials, tests, and are capable of initializing the interaction with learners by providing advice, necessary instant messages, etc., to online learners. One of the major challenges involved in developing personalized VLEs is to achieve effective personalization functionalities, such as personalized content management, learner model, learner plan and adaptive instant interaction. Autonomous intelligent agents provide an important technology for accomplishing personalization in VLEs. A number of agents work collaboratively to enable personalization by recognizing an individual's eLeaming pace and reacting correspondingly. In this research, a personalization model has been developed that demonstrates dynamic eLearning processes; secondly, this study proposes an architecture for PVLE by using intelligent decision-making agents' autonomous, pre-active and proactive behaviors. A prototype system has been developed to demonstrate the implementation of this architecture. Furthemore, a field experiment has been conducted to investigate the performance of the prototype by comparing PVLE eLearning effectiveness with a non-personalized VLE. Data regarding participants' final exam scores were collected and analyzed. The results indicate that intelligent agent technology can be employed to achieve personalization in VLEs, and as a consequence to improve eLeaming effectiveness dramatically.

A general-purpose tunable landscape generator

The research literature on metalieuristic and evolutionary computation has proposed a large number of algorithms for the solution of challenging real-world optimization problems. It is often not possible to study theoretically the performance of these algorithms unless significant assumptions are made on either the algorithm itself or the problems to which it is applied, or both. As a consequence, metalieuristics are typically evaluated empirically using a set of test problems. Unfortunately, relatively little attention has been given to the development of methodologies and tools for the large-scale empirical evaluation and/or comparison of metaheuristics. In this paper, we propose a landscape (test-problem) generator that can be used to generate optimization problem instances for continuous, bound-constrained optimization problems. The landscape generator is parameterized by a small number of parameters, and the values of these parameters have a direct and intuitive interpretation in terms of the geometric features of the landscapes that they produce. An experimental space is defined over algorithms and problems, via a tuple of parameters for any specified algorithm and problem class (here determined by the landscape generator). An experiment is then clearly specified as a point in this space, in a way that is analogous to other areas of experimental algorithmics, and more generally in experimental design. Experimental results are presented, demonstrating the use of the landscape generator. In particular, we analyze some simple, continuous estimation of distribution algorithms, and gain new insights into the behavior of these algorithms using the landscape generator.

A multiresolution terrain model for efficient visualization query processing

Multiresolution Triangular Mesh (MTM) models are widely used to improve the performance of large terrain visualization by replacing the original model with a simplified one. MTM models, which consist of both original and simplified data, are commonly stored in spatial database systems due to their size. The relatively slow access speed of disks makes data retrieval the bottleneck of such terrain visualization systems. Existing spatial access methods proposed to address this problem rely on main-memory MTM models, which leads to significant overhead during query processing. In this paper, we approach the problem from a new perspective and propose a novel MTM called direct mesh that is designed specifically for secondary storage. It supports available indexing methods natively and requires no modification to MTM structure. Experiment results, which are based on two real-world data sets, show an average performance improvement of 5-10 times over the existing methods.

Hierarchical indexing structure for efficient similarity search in video retrieval

With the rapid increase in both centralized video archives and distributed WWW video resources, content-based video retrieval is gaining its importance. To support such applications efficiently, content-based video indexing must be addressed. Typically, each video is represented by a sequence of frames. Due to the high dimensionality of frame representation and the large number of frames, video indexing introduces an additional degree of complexity. In this paper, we address the problem of content-based video indexing and propose an efficient solution, called the Ordered VA-File (OVA-File) based on the VA-file. OVA-File is a hierarchical structure and has two novel features: 1) partitioning the whole file into slices such that only a small number of slices are accessed and checked during k Nearest Neighbor (kNN) search and 2) efficient handling of insertions of new vectors into the OVA-File, such that the average distance between the new vectors and those approximations near that position is minimized. To facilitate a search, we present an efficient approximate kNN algorithm named Ordered VA-LOW (OVA-LOW) based on the proposed OVA-File. OVA-LOW first chooses possible OVA-Slices by ranking the distances between their corresponding centers and the query vector, and then visits all approximations in the selected OVA-Slices to work out approximate kNN. The number of possible OVA-Slices is controlled by a user-defined parameter delta. By adjusting delta, OVA-LOW provides a trade-off between the query cost and the result quality. Query by video clip consisting of multiple frames is also discussed. Extensive experimental studies using real video data sets were conducted and the results showed that our methods can yield a significant speed-up over an existing VA-file-based method and iDistance with high query result quality. Furthermore, by incorporating temporal correlation of video content, our methods achieved much more efficient performance.

Approximate processing of massive continuous quantile queries over high-speed data streams

Quantile computation has many applications including data mining and financial data analysis. It has been shown that an is an element of-approximate summary can be maintained so that, given a quantile query d (phi, is an element of), the data item at rank [phi N] may be approximately obtained within the rank error precision is an element of N over all N data items in a data stream or in a sliding window. However, scalable online processing of massive continuous quantile queries with different phi and is an element of poses a new challenge because the summary is continuously updated with new arrivals of data items. In this paper, first we aim to dramatically reduce the number of distinct query results by grouping a set of different queries into a cluster so that they can be processed virtually as a single query while the precision requirements from users can be retained. Second, we aim to minimize the total query processing costs. Efficient algorithms are developed to minimize the total number of times for reprocessing clusters and to produce the minimum number of clusters, respectively. The techniques are extended to maintain near-optimal clustering when queries are registered and removed in an arbitrary fashion against whole data streams or sliding windows. In addition to theoretical analysis, our performance study indicates that the proposed techniques are indeed scalable with respect to the number of input queries as well as the number of items and the item arrival rate in a data stream.