By Topic

Unsupervised Learning via Self-Organization:A Dynamic Approach

Cover Image Copyright Year: 2014
Author(s): Kyan, M.; Muneesawang, P.; Jarrah, K.; Guan, L.
Publisher: Wiley-IEEE Press
Content Type : Books & eBooks
Topics: Computing & Processing (Hardware/Software)
  • Print
  •   Click to expandTable of Contents

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Front Matter

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.fmatter
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      The prelims comprise:
      Half Title
      Editorial Board
      Title
      Copyright
      Contents
      Acknowledgments View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Introduction

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch1
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      Automation requires computational systems that exhibit some degree of intelligence, in terms of the ability of a system to formulate its own models of the data in question with little or no user intervention. This introductory chapter provides an overview of the content discussed in the subsequent chapters of the book. The book primarily introduces a new approach to the general problem of unsupervised learning, based on the principles of dynamic self-organization. It gives an extensive review of the general problems of unsupervised clustering, with emphasis placed on the inherent relationship that exists between unsupervised learning and self-organization. The book presents self-organizing tree map (SOTM) and its recently successful application in multimedia processing. It describes the developments of the self-organizing hierarchical variance map (SOHVM) and its application in the unsupervised segmentation and visualization of microbiological image data. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Unsupervised Learning

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch2
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      In this chapter, a general review of Unsupervised Learning is conducted. Generic clustering issues are first defined and explained. A survey of traditional approaches to Unsupervised Learning is then presented, and the chapter concludes in with a discussion of assessment measures and limitations in the evaluation of clustering solutions. It presents a brief survey of the issues that need to be considered in assessing the validity of unsupervised clustering results. Distance metrics, cluster quality, and cluster validity are each vast topics unto themselves and become essential, yet difficult considerations in the evaluation of a clustering solution within unsupervised contexts. These are expanded upon in the chapter. It outlines some of the more popular clustering approaches from the literature namely, Iterative Mean-Squared Error Approaches, Mixture Decomposition Approaches, Agglomerative Hierarchical Approaches, Graph-Theoretic Approaches, Evolutionary Approaches and Neural Network Approaches. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Self-Organization

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch3
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      This chapter presents the principles of Self-Organization, and focuses on Adaptive Resonance Theory (ART) and Self-Organizing Map (SOM) neural networks. It investigates the theoretic basis of formulations of these neural networks, and illustrates a few examples. The structures of these networks and their learning algorithms are also thoroughly explored in the chapter. The ART architecture is a specifically designed neural network to overcome the stability-plasticity dilemma. It is described using nonlinear differential equations. In addition to ART and SOM, there are two other fixed approaches that could be considered fundamental, namely, Neural Gas and the Hierarchical Feature Map. While both are strongly related to the SOM in terms of the learning mechanism, they each have spawned a range of newer architectures are introduced in the chapter. The chapter explains other popular architectures to have emerged based on similar principles of Self-Organization, drawing a distinction between static and dynamic architectures. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Self-Organizing Tree Map

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch4
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      This chapter proposes a new mechanism named the self-organizing tree map (SOTM). The motivation of the new method is twofold: (a) to keep the ART's ability to create new output neurons dynamically while overcoming the Global threshold setting problem; (b) to keep the SOM's property of topology preservation while strengthening the flexibility of adapting to changes in the input distribution and maximally reflecting the distribution of the input patterns. In the SOTM, relationships between the output neurons can be dynamically defined during learning. There are thus, two different levels of adaptation in the SOTM, which involves weight adaptation and structure adaptation. The basic principle underlying competitive learning is vector quantization. The chapter demonstrates dynamic topology and classification capability are two prominent characteristics of the SOTM. Finally, the SOTM model not only enhanced the ART's autonomous category classification and the SOM's topology preservation, but also overcame their weaknesses. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Self-Organization in Impulse Noise Removal

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch5
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      This chapter proposes a novel approach for suppressing impulse noise in digital images while effectively preserving more details than previously proposed methods. The method presented is based on impulse noise detection and noise exclusive restoration. A noise-exclusive median (NEM) filtering algorithm and a noise-exclusive arithmetic mean (NEAM) filtering algorithm are proposed to restore the image. The chapter proposes a more general model in which a noisy pixel can take on arbitrary values in the dynamic range. It presents an iterative approach that is introduced using a recursive NEM filter to restore images corrupted by very strong noise. The chapter proposes a new filtering scheme, which aimed to keep the robustness of median-type and nonlinear mean filters while overcoming their disadvantages. It performs experiments based on different models and with varying percentage of noise. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Self-Organization in Image Retrieval

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch6
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      This chapter provides a comprehensive study on modern approaches in the area of image indexing and retrieval on the use of Self-Organization as a core enabling technology. It begins with the development of Content-based image retrieval (CBIR) systems, which includes the implementation of a radial basis function (RBF) based relevance feedback (RF) method. The chapter presents automatic and semiautomatic methods in multimedia retrieval, using the pseudo-RF for minimizing user interaction in a retrieval process. It introduces a framework for a novel extension of the self-organizing tree map (SOTM) for hierarchical clustering, the Directed SOTM (DSOTM). It demonstrates an optimized architecture for an automatic retrieval system based on collaboration between the DSOTM and the Genetic Algorithm (GA). A study on the feasibility of the proposed feature weight detection scheme in conjunction with the DSOTM, SOTM, and self-organizing feature map (SOFM) classifier techniques is presented. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      The Self-Organizing Hierarchical Variance Map

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch7
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      This chapter introduces and develops a new model for clustering data, offering a number of enhancements and features over the self-organizing tree map (SOTM). The model is known as the Self-Organizing Hierarchical Variance Map (SOHVM). The chapter highlights some of its limitations. In so doing, a motivation is provided for a more advanced clustering algorithm, one that retains some of the desirable properties of the SOTM. The component responsible for mapping local variance information is known as a Hebbian Maximal Eigenfilter (HME). It outlines and justifies the key components and principles of operation for the new model. In addition, the implementation details are discussed. Finally, a series of visual simulations on synthetic two-dimensional (2D) data are presented, with the goal of providing a simple and clear demonstration of the new model in operation, highlighting some of its key features and strengths over popular existing architectures from the literature. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Microbiological Image Analysis Using Self-Organization

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch8
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      This chapter considers the potential and flexibility of self-organizing tree map (SOTM) based and self-organizing hierarchical variance map (SOHVM) based learning for tasks in microbiological image analysis. As a demonstration of the SOHVM's ability to mine topological information from an input space, the chapter describes with an example for how such information can be used to simplify the task of visualizing a large three-dimensional (3D) stack of phase-contrast acquired plant chromosomes imaged during an advanced state of mitosis (cell division). The chapter considers two types of microbiological image data in order to demonstrate the potential for the proposed algorithm to achieve unsupervised, fully automatic segmentations. It shows examples of utilizing this automated property of the SOHVM to seek more natural segmentations of gray-level and higher order, multidimensional feature descriptions, with examples for the clustering of texture information and Local gray-level-based statistics. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Closing Remarks and Future Directions

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.ch9
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      The focal points of the book lay in the design and development of two novel models for unsupervised learning or data clustering, based on dynamic Self-Organization: namely, the self-organizing tree map (SOTM) and The Self-Organizing Hierarchical Variance Map (SOHVM). This chapter summarizes the main properties and recommendations in the use of such models, and discusses some potential directions for future research and application. The real advantage of creating a self-organized clustering as opposed to most other clustering methods, lies in the availability of the resulting topological map. Mining the topology, as opposed to assuming one through imposing predetermined lattice, can be leveraged for very specific tasks. The chapter focuses on three major categories of task: namely, dynamic navigation through information repositories; knowledge-assisted visualization; and path-based trajectory analysis. In each category, there is a common Theme-where there is topology, there is context, and context can assist in conveying or extracting knowledge. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Appendix A

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.app1
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      No Abstract. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      References

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.refs
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      No Abstract. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      Index

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.index
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      No Abstract. View full abstract»

    • Full text access may be available. Click article title to sign in or learn about subscription options.

      IEEE Press Series on Computational Intelligence

      Kyan, M. ; Muneesawang, P. ; Jarrah, K. ; Guan, L.
      Unsupervised Learning via Self-Organization:A Dynamic Approach

      DOI: 10.1002/9781118875568.oth
      Copyright Year: 2014

      Wiley-IEEE Press eBook Chapters

      No Abstract. View full abstract»