Applications of Soft Computing
Recent Trends
1st Edition
Published on 8. April 2010
XIX, 420 pages
978-3-540-36266-1 (ISBN)
Description
Soft Computing is a complex of methodologies that embraces approximate reasoning, imprecision, uncertainty and partial truth in order to mimic the remarkable human capability of making decisions in real-life, ambiguous environments. Soft Computing has therefore become popular in developing systems that encapsulate human expertise. Applications of Soft Computing: Recent Trends contains a collection of papers that were presented at the 10th Online World Conference on Soft Computing in Industrial Applications, held in September 2005. This carefully edited book provides a comprehensive overview of the recent advances in the industrial applications of soft computing and covers a wide range of application areas, including optimisation, data analysis and data mining, computer graphics and vision, prediction and diagnosis, design, intelligent control, and traffic and transportation systems. The book is aimed at researchers and professional engineers who are engaged in developing and applying intelligent systems. It is also suitable as wider reading for science and engineering postgraduate students.
More details
Other editions
Additional editions
Book
06/2006
Springer
€213.99
Shipment within 7-9 days
Content
Computer Graphics, Imaging and Vision.- Efficient Genetic Algorithms for Arabic Handwritten Characters Recognition.- Generic Black Box Optimisation Algorithms for Colour Quantisation.- Neural Network Combined with Fuzzy Logic to Remove Salt and Pepper Noise in Digital Images.- Computing Optimized NURBS Curves using Simulated Evolution on Control Parameters.- Control and Robotics.- Design of A Takagi-Sugeno Fuzzy Compensator for Inverted Pendulum Control Using Bode Plots.- Soft Computing in Accuracy Enhancement of Machine Tools.- Mobile Robot Navigation: Potential Field Approach Vs. Genetic-Fuzzy System.- Intelligent Tuning and Application of a PID Controller Using Universal Model.- Design.- Dynamic Reconfiguration Algorithm for Field Programmable Analog Scalable Device Array (FPADA) with Fixed Topology.- Advances in the Application of Machine Learning Techniques in Drug Discovery, Design and Development.- A Review on Design Optimisation and Exploration with Interactive Evolutionary Computation.- Pattern Recognition.- Mapping of Natural Patterns by Liquid Architectures Implementing Neural Cliques.- Pattern Recognition Using Modular Neural Networks and Fuzzy Integral as Method for Response Integration.- Genetic Algorithm-Evolved Bayesian Network Classifier for Medical Applications.- A Real-Time Hand Gesture Interface for Medical Visualization Applications.- Classification.- A Hybrid Intelligent System and Its Application to Fault Detection and Diagnosis.- Evolutionary Multidimensional Scaling for Data Visualization and Classification.- Extended Genetic Algorithm for Tuning a Multiple Classifier System.- Development of Fuzzy Expert System for Customer and Service Advisor Categorisation within Contact Centre Environment.- Soft Computing for Intelligent Information Management.- Soft Computing in Intelligent Data Analysis.- Identification and Forecasting.- Neural Network-Based Expert System to Predict the Results of Finite Element Analysis.- Modular Neural Networks with Fuzzy Integration Applied to Time Series Prediction.- Fuzzy Model Identification for Rapid Nickel-Cadmium Battery Charger through Particle Swarm Optimization Algorithm.- Fuzzy Association Rule Mining for Model Structure Identification.- Modeling Public Transport Trips with General Regression Neural Networks; A Case Study for Istanbul Metropolitan Area.- Optimization Problems: Assignment, Partitioning and Ordering.- Use of Genetic Algorithm to Optimum Test Frequency Selection.- Performance Analysis of Parallel Strategies for Bi-objective Network Partitioning.- A Guided Rule Reduction System for Prioritization of Failures in Fuzzy FMEA.- A Hybrid Method of Differential Evolution and SQP for Solving the Economic Dispatch Problem with Valve-Point Effect.- Multiobjective Prioritization in the Analytic Hierarchy Process by Evolutionary Computing.- Evolutionary and Heuristic Algorithms for Multiobjective 0-1 Knapsack Problem.- The Assignment of Referees to WSC10 Submissions: An Evolutionary Approach.- Optimization Methods: Development and Analysis.- Robustness using Multi-Objective Evolutionary Algorithms.- Genetic Programming, Probabilistic Incremental Program Evolution, and Scalability.- Adaptive Parameter Control of Evolutionary Algorithms Under Time Constraints.- Role of Chaos in Swarm Intelligence - A Preliminary Analysis.- Multi-parent Recombination Operator with Multiple Probability Distribution for Real Coded Genetic Algorithm.- Tutorials.- Special Tutorial - State of the Art Face Detection: Cascade Boosting Approaches.- Special Tutorial - Particle Swarms for FuzzyModels Identification.- Special Tutorial - Project Management: Issues in Computer Based Monitoring and Control with Soft Computing Approaches.
"Evolutionary Multidimensional Scaling for Data Visualization and Classification (p. 177-178)
Summary. Multidimensional Scaling (MDS) is a well established technique for the projection of high-dimensional data in pattern recognition, data visualization and analysis, as well as scientific and industrial applications. In particular, Sammons Nonlinear Mapping (NLM) as a common MDS instance, computes distance preserving mapping based on gradient descent, which depends on initialization and just can reach the nearest local optimum. Improvement of mapping quality or reduction of mapping error is aspired and can be achieved by more powerful optimization techniques, e.g., stochastic search, successfully applied in our prior work. In this paper, evolutionary optimization adapted to the particular problem and the NLM has been investigated for the same aim, showing the feasibility of the approach and delivering competitive and encouraging results.
1 Introduction
Multidimensional Scaling (MDS) is a common dimensional reduction method and widely used in exploratory data analysis and design of pattern recognition and integrated sensor systems for extraction of essential information from multivariate data sets [1]. Extraction of essential information by dimensionality reduction is required for various reasons. For instance, it can avoid the curse of dimensionality and thus improve the ability of classification. The analysis of unknown data by projection and ensuing interactive visualization is another common and important application of dimensionality reduction [2, 3].
MDS represents data in a smaller number of dimensions preserving the similarity structure of the data as much as possible. One common MDS instance is Sammons nonlinear mapping (NLM) that will be emphasized and discussed in this paper. In this method, a criterion denoted stress (error) function is defined and optimization takes place by gradient descent techniques [4]. These optimizations have known drawbacks, which strongly depend on the initialization, can get more easily trapped in a less fortunate local minimum and saturate on an undesirable high error value.
Implications of this behavior can be data visualizations of low reliability, which incorporate twists and misleading neighborhoods. To overcome this problem, there are some methods that have been proposed, e.g., simulated annealing [5] or using neural network [6]. In similar cases and applications, stochastic search optimization in our prior work has been successfully applied to overcome such a problem [7]. In this paper, we investigated another stochastic method, that draws inspiration from the process of natural evolution, i.e., evolutionary computation. In several research activities, MDS has been applied to extract useful information and visualize the progress of evolutionary optimization [8].
Additionally, some recent publications have introduced the application of genetic algorithm for optimizing MDS itself [9, 10]. Similar to those, we have applied evolutionary optimization. However, we adopted from our work on stochastic search [7] the concept of the mutation operator, rather than employing gradient factor, since this approach pursued in [9] takes significant computational effort. Regarding to the goals of mapping improvement, i.e., the reliability (error reduction) and speed computation, we examine the feasibility and assessment of applicability or competitiveness of such approaches adapted to MDS."
Summary. Multidimensional Scaling (MDS) is a well established technique for the projection of high-dimensional data in pattern recognition, data visualization and analysis, as well as scientific and industrial applications. In particular, Sammons Nonlinear Mapping (NLM) as a common MDS instance, computes distance preserving mapping based on gradient descent, which depends on initialization and just can reach the nearest local optimum. Improvement of mapping quality or reduction of mapping error is aspired and can be achieved by more powerful optimization techniques, e.g., stochastic search, successfully applied in our prior work. In this paper, evolutionary optimization adapted to the particular problem and the NLM has been investigated for the same aim, showing the feasibility of the approach and delivering competitive and encouraging results.
1 Introduction
Multidimensional Scaling (MDS) is a common dimensional reduction method and widely used in exploratory data analysis and design of pattern recognition and integrated sensor systems for extraction of essential information from multivariate data sets [1]. Extraction of essential information by dimensionality reduction is required for various reasons. For instance, it can avoid the curse of dimensionality and thus improve the ability of classification. The analysis of unknown data by projection and ensuing interactive visualization is another common and important application of dimensionality reduction [2, 3].
MDS represents data in a smaller number of dimensions preserving the similarity structure of the data as much as possible. One common MDS instance is Sammons nonlinear mapping (NLM) that will be emphasized and discussed in this paper. In this method, a criterion denoted stress (error) function is defined and optimization takes place by gradient descent techniques [4]. These optimizations have known drawbacks, which strongly depend on the initialization, can get more easily trapped in a less fortunate local minimum and saturate on an undesirable high error value.
Implications of this behavior can be data visualizations of low reliability, which incorporate twists and misleading neighborhoods. To overcome this problem, there are some methods that have been proposed, e.g., simulated annealing [5] or using neural network [6]. In similar cases and applications, stochastic search optimization in our prior work has been successfully applied to overcome such a problem [7]. In this paper, we investigated another stochastic method, that draws inspiration from the process of natural evolution, i.e., evolutionary computation. In several research activities, MDS has been applied to extract useful information and visualize the progress of evolutionary optimization [8].
Additionally, some recent publications have introduced the application of genetic algorithm for optimizing MDS itself [9, 10]. Similar to those, we have applied evolutionary optimization. However, we adopted from our work on stochastic search [7] the concept of the mutation operator, rather than employing gradient factor, since this approach pursued in [9] takes significant computational effort. Regarding to the goals of mapping improvement, i.e., the reliability (error reduction) and speed computation, we examine the feasibility and assessment of applicability or competitiveness of such approaches adapted to MDS."
