This work is based on a diploma thesis by Changxing Dong [Dong04]. In [Dong04], the effects of different shapespaces on the behaviour of artificial immune system algorithms are evaluated. For that purpose a pattern recognition task was used. In this case, chinese characters should be learned and recognized. Each character was adapted individually using theCLONALGalgorithm [dCT02]. The resulting set of antibody patterns, the duration of running the algorithm and the quality of recognition were then used to evaluate the differences between the different shapespaces.
Inhaltsverzeichnis (Table of Contents)
- Introduction
- Background
- Fundamental Ideas
- The Core Immune Learning Algorithm
- Problem Description
- Deciding the Immune Principles to be used for Problem Solving.
- Engineering the Artificial Immune System
- Defining the Types of Immune Components to be used
- Defining the Mathematical Representation for the Elements of the AIS
- Applying Immune Principles to Problem Solving
- The Metadynamics of the AIS
- Reverse Mapping from AIS to the Real Problem
- The Ra Algorithms
- The Full Learning Algorithm in Pseudocode
- The aiNet Algorithm.
- The Isis Algorithm
- Comparison of aiNet and Isis
- The Architecture
- The Concept
- The Ra Algorithms
- The Learning Algorithm: ra Create
- The Classification Algorithm: raCluster
- Implementation
- Implemented Classes
- Invoking the Algorithms
- Isis.
- raCreate
- raCluster
- raGraph.
- Data Preprocessing
- Results
- Results with the Identity Shapespace.
- Level 4 Clusters with Identity Shapespace
- Comments.
- Results with the Convolution Shapespace
- Level 2 Clusters with Convolution Shapespace
- Comments .
- Results with the Fourier Shapespace
- Level 1 Clusters with Fourier Shapespace
- Comments.
- Comparison of the Algorithm Behaviour Across the Different Shapespaces.
- Further Possibilities for the Ra/Isis Approach
- Parametertuning
- Hierarchical Parameters
- Shapespaces.
- Hierarchical Shapespaces
- Clustering Unusual Data.
- Data Analysis and Other Learning Algorithms
- Classification
- Hierarchical Classification
- Classification Networks
- Subsequent Classification
- Analysis of the Algorithm Behaviour
- Ra/Isis vs. Common Aglomerative Clustering
- Comparison of Ra/Isis with Self Organizing Maps
- Results of Applying the Data to a SOM
Zielsetzung und Themenschwerpunkte (Objectives and Key Themes)
This paper presents a novel clustering algorithm that utilizes an artificial immune system. The algorithm aims to create a conceptual hierarchy by iteratively applying the immune system algorithm on its own output. The paper explores the architectural setup for recurrent immune systems, outlining its implementation and presenting experimental findings. The system's performance is compared to other clustering algorithms, establishing analogies between artificial immune systems, self-organizing maps, and similar methods.
- Development and implementation of a clustering algorithm based on an artificial immune system.
- Creation of a conceptual hierarchy by iteratively applying the algorithm on its output.
- Architectural setup for recurrent immune systems.
- Comparison of the system's performance with other clustering algorithms.
- Exploration of analogies between artificial immune systems, self-organizing maps, and similar methods.
Zusammenfassung der Kapitel (Chapter Summaries)
The paper starts by introducing the background and fundamental ideas behind the research. It then delves into the core immune learning algorithm, outlining the problem description, the selection of immune principles for problem solving, and the engineering of the artificial immune system. The chapter also details the metadynamics of the AIS, the reverse mapping from AIS to the real problem, and presents the full learning algorithm in pseudocode. The Ra algorithms, specifically aiNet and Isis, are introduced, highlighting their comparison and algorithmic structure.
Chapter 3 focuses on the architecture of the system, explaining the concept of recurrent immune systems and the Ra algorithms, including the learning algorithm (ra Create) and the classification algorithm (raCluster). Chapter 4 discusses the implementation, highlighting the implemented classes and the process of invoking the algorithms, including Isis, raCreate, raCluster, and raGraph. Chapter 5 delves into the data preprocessing techniques used in the research.
Chapter 6 presents the results obtained with the Identity Shapespace, Convolution Shapespace, and Fourier Shapespace, including comments on each shapespace. The chapter concludes by comparing the algorithm's behavior across the different shapespaces.
Chapter 7 explores further possibilities for the Ra/Isis approach, including parameter tuning, hierarchical parameters, hierarchical shapespaces, clustering unusual data, data analysis and other learning algorithms, and classification techniques like hierarchical classification, classification networks, and subsequent classification.
Chapter 8 analyzes the algorithm behavior, comparing Ra/Isis with common agglomerative clustering and self-organizing maps.
Schlüsselwörter (Keywords)
Artificial immune system, clustering algorithm, recurrent immune system, conceptual hierarchy, self-organizing map, shapespace, data analysis, classification, machine learning.
- Quote paper
- Stefan Schadwinkel (Author), 2005, A Recurrent Architecture of Independent Single Immune Systems, Munich, GRIN Verlag, https://www.grin.com/document/48178
