Skip to main content
Fachgebiete
Automobil + Motoren Bauwesen + Immobilien Business IT + Informatik Elektrotechnik + Elektronik Energie + Nachhaltigkeit Finance + Banking Management + Führung Marketing + Vertrieb Maschinenbau + Werkstoffe Versicherung + Risiko
DE
EN
Bücher
Zeitschriften
Themenseiten
Organisationspsychologie Projektmanagement Marketing Smart Manufacturing
Weitere Formate
Podcasts Webinare Technik Webinare Wirtschaft Kongresse Veranstaltungskalender Awards
Jetzt Einzelzugang starten
Zugang für Unternehmen
Referenzkunden
SLX-Digitalkonferenz: Zukunftswerkstatt 2024

Mehr Umsatz mit Top-Kontakten

Am 7. Mai erfahren Sie, wie Vertriebsteams Leadgenerierung, Leadnurturing und Leadstrategien effizient steuern können.
Erweiterte Suche
Anmelden
nach oben

2010 | Buch

Introduction Kapitel lesen Erstes Kapitel lesen

Towards Hybrid and Adaptive Computing

A Perspective

verfasst von: Anupam Shukla, Ritu Tiwari, Rahul Kala

Verlag: Springer Berlin Heidelberg

Buchreihe : Studies in Computational Intelligence

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten
insite
SUCHEN

Über dieses Buch

Soft Computing today is a very vast field whose extent is beyond measure. The boundaries of this magnificent field are spreading at an enormous rate making it possible to build computationally intelligent systems that can do virtually anything, even after considering the hostile practical limitations. Soft Computing, mainly comprising of Artificial Neural Networks, Evolutionary Computation, and Fuzzy Logic may itself be insufficient to cater to the needs of various kinds of complex problems. In such a scenario, we need to carry out amalgamation of same or different computing approaches, along with heuristics, to make fabulous systems for problem solving. There is further an attempt to make these computing systems as adaptable as possible, where the value of any parameter is set and continuously modified by the system itself. This book first presents the basic computing techniques, draws special attention towards their advantages and disadvantages, and then motivates their fusion, in a manner to maximize the advantages and minimize the disadvantages. Conceptualization is a key element of the book, where emphasis is on visualizing the dynamics going inside the technique of use, and hence noting the shortcomings. A detailed description of different varieties of hybrid and adaptive computing systems is given, paying special attention towards conceptualization and motivation. Different evolutionary techniques are discussed that hold potential for generation of fairly complex systems. The complete book is supported by the application of these techniques to biometrics. This not only enables better understanding of the techniques with the added application base, it also opens new dimensions of possibilities how multiple biometric modalities can be fused together to make effective and scalable systems.

Inhaltsverzeichnis

Frontmatter

Simple Computing Techniques

Frontmatter
Introduction
Abstract
Soft Computing Systems have undertaken a radical change largely attributed to their widespread use along with a vast research community that has developed over the years. Before beginning the ballad of development of this domain, it is important to get the stage set. This chapter explores the various concepts and terms used in computationally intelligent systems of today. We give a brief introduction to recognition systems, machine learning, expert systems and biometric identification. The major focus of the chapter is upon presenting the application of Soft Computing systems, the manner in which Soft Computing approaches contribute towards the application, and the various problems and issues that the application presents. These issues open gateways for a lot of research for the research community. While the sophisticated Soft Computing systems of today may be able to effectively solve a wide variety of problems, the data availability and computational constraints would always be a limitation for the flawless growth of the soft computing systems.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Artificial Neural Networks
Abstract
Artificial Neural Networks (ANN) are an inspiration from the human brain. These systems contain a large number of neurons that work in a parallel architecture. Each neuron takes its input directly from system or from other neurons. The information is processed and given to the other neurons. This is the basic phenomenon that makes possible all simple and complex problem solving ability of these networks. The chapter discusses the various models of neural networks that include multi-layer perceptron with back propagation algorithm, radial basis function networks, learning vector quantization, self organizing maps and recurrent neural networks. We discuss the basic philosophies and problem solving approach of these networks. A lot of emphasis is given on the various system parameters and their role and importance in the overall system design. We further illustrate the various limitations of the different models. This forms the motivation behind the use of hybrid systems that we present in the subsequent chapters.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Genetic Algorithm
Abstract
One of the most exciting aspects of life is its evolutionary nature where the individuals keep improving along with generations. Genetic Algorithms are an inspiration from this natural evolution and find themselves as powerful optimizing agents for solving numerous real life applications. These algorithms can model complex problems and return the optimal solution in an iterative manner. This chapter presents the manner in which we model and solve the problem using this evolutionary technique. The role of the various parameters and optimal parameter setting as per the problem requirements would be discussed. The chapter would present mutation, selection, crossover and other genetic operators. Evolution forms the base for most of the complex systems that are designed to evolve with time. In this chapter we hence first study the basic concepts and then take an inspiration towards evolving systems. At the same time we present the limitations of evolution that marks a threshold to massive potential of problem solving that these algorithms have.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Fuzzy Logic
Abstract
Logic forms a fundamental concept behind the manner in which a variety of works are performed. One tries to process the inputs based on reasoning or logic that helps in the generation of the outputs. In this chapter we first study fuzzy logic that is driven by fuzzy sets and fuzzy rules. Then we present the Fuzzy Inference Systems that use Fuzzy Logic for the generation of the outputs from the inputs. The logic base of these systems gives a clear understanding of the manner in which the system is operating. The chapter would present the various concepts of problem solving using Fuzzy Logic covering membership functions, fuzzy arithmetic and operators and finally the fuzzy inference systems. We also present the analogy of these systems with the Artificial Neural Networks and the design methodology in fuzzy systems. At the end the chapter focuses on the problems and limitations of the fuzzy systems that motivate their hybridization in form of hybrid computing systems that give a better performance.
Anupam Shukla, Ritu Tiwari, Rahul Kala

Simple Intelliegent Systems

Frontmatter
Speech Signal Analysis
Abstract
Intelligent systems possess the capability to model and solve many problems of practical importance. The best way to understand these systems is do design and develop such systems which exposes their various advantages and disadvantages. This chapter presents the basic analysis technique of speech signals that would further help us in using speech as a medium of developing intelligent systems. In this chapter we study the manner in which we may highlight and extract useful features out of a given speech signals. We discuss the analysis techniques in two heads. The first head consist of the bank of filters approach. Here we present the Fourier, Short Time Fourier and Wavelet Analysis which extract interesting features. Here we would stress the importance of time and frequency domain. In the other head we would discuss the Linear Predictive Coding (LPC) methods. We discuss the manner in which the linear coding helps in analysis. We even discuss the general speech parameters that facilitate good recognition in these intelligent systems.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Intelligent System Design in Speech
Abstract
Speech based systems may be used for a variety of tasks that we use in everyday life. The applications using speech may be broadly classified into speaker recognition and speaker verification. The recognition deals with the identification of the person who is speaking. The task of verification systems is to verify the claimed identity of the person. Here we make use of speech as a biometric modality. Besides, speech is extensively used for word identification whose application includes the speech to text conversion. The variety of applications has a large number of issues associated with them. In this chapter we give a broad overview of these systems, their working and design issues. Some of the issues addressed in the chapter include the variation of speech, effect of noise, text dependent and independent recognition, start and end of signal detection, etc. Most of the discussions revolve around the pattern recognition approach for the recognition systems. This would give an application base of the discussed approaches that we have studied so far or would be studying later in this book.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Intelligent Systems Design in Music
Abstract
Music is an exciting area where the computational intelligence has cast a deep impact. In this chapter we study a few intelligent systems in the domain of music. The vast volume of music available in various formats has necessitated the need for their automated classification. Here we discuss the system for the identification of genres as well as the recognition of artists. These systems have a variety of application in playlist generation, music suggestion, music fetching etc. The other part of the chapter would focus upon the composition of music. Here also we discuss a variety of methods using Genetic Algorithms and Neural Networks. The manual assistive design of Genetic Algorithms enables the automated composition of music as per human demand. The neural approach uses a series prediction phenomenon to compose music when some part of it is known. These systems enable good composition techniques which may be employed to assist human composers in their task.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Intelligent System Design in Face
Abstract
Speech forms a major biometric that makes it possible to reap the numerous fruits of automation and verification. The later of the book focus on the shortcomings of speech that motivate the development of hybrid systems using face along with speech. Hence it is important to have a brief idea about face as a biometric that uniquely couples with speech. This chapter would present various techniques of using face as a biometric. We deal with two kinds of systems. The first class of systems uses dimensionality reduction techniques over the visual information that face represents. Here we would discuss the Principle Component Analysis (PCA) and Regularized Linear Discriminant Analysis (R-LDA). All these methods have their own mechanisms for dimensionality reduction. The other class of algorithm identifies landmarks in face and uses these for the recognition purpose. We also analyze the various advantages and shortcomings of these systems.
Anupam Shukla, Ritu Tiwari, Rahul Kala

Evolutionary Computing

Frontmatter
Swarm Intelligence
Abstract
The manner in which the swarms behave and survive is a great inspiration to build evolutionary algorithms. The flocking of birds, foraging of ants, etc. has been a big inspiration for the optimization and search algorithms that we study in this chapter. We first present the Particle Swarm Optimization which is an optimization tool inspired by the flocking of birds. We present the manner in which multiple particles walk on the fitness landscape to search the optima. We then present the Ant Colony Algorithm which is a search based algorithm inspired from the working of the ant colonies. We would present the manner in which the ants proceed to search the most optimal solution and help the others by the deposition of the pheromone. The other algorithm for discussion is the bee algorithm inspired by the honey bees. We also present Stochastic Diffusion Search, Probability Based Incremental Learning and Biogeography Based Optimization in this chapter.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Genetic Programming
Abstract
Evolution is becoming a comprehensive tool for automated problem solving that enables the systems to attain their optimal form. Evolution includes the three basic methodologies of Genetic Algorithm, Genetic Programming and Evolutionary Strategies. In this chapter we focus our attention towards the use of Genetic Programming as an evolution tool where programs to solve the problem are evolved along with time. We state the working of these algorithms with a keen eye on their differences with the Genetic Algorithm. Selection, Mutation and Crossover enable the generation of one population to other. A major problem in their use is the unnecessary growth of the code along with the generations which would be addressed in the chapter. Later we present a specialized form of Genetic Programming called the Grammatical Evolution that uses a set of Grammatical Rules running by Backus-Naur Form to evolve systems. This forms another great class of algorithms for optimization or evolution.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Evolutionary Strategies
Abstract
Genetic Algorithms with their complex genetic operators make a complex system that is time consuming and difficult to comprehend. Most of the systems we encounter in everyday life may not require that complex formulation. In this chapter we study a major constituent of the Evolutionary Algorithm i.e. Evolutionary Strategies. The chapter would focus upon the manner in which this problem solving technique tackles the problem with much less individuals and simple genetic operators. We further discuss the adaptation of the operators that this technique employs. The chapter especially would focus upon the adaptive nature of the Evolutionary Strategies. Here the individuals assess the fitness landscape and make their move accordingly. The general framework of Evolutionary Strategies generates λ individuals from a population of μ individuals with each individual resulting from crossover by ρ parents. Each individual may survive for k generations. The chapter explores the various concepts, issues and applications of this problem solving methodology.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Other Evolutionary Concepts
Abstract
Evolutionary Algorithms in numerous forms present powerful problem solving tools. These tools have been modified and adapted to various kinds of problems as per the requirements. Genetic Algorithms, Evolutionary Strategies and Genetic Programming happen to be the basic classifications of these algorithms. In this chapter we further discuss some of the widely used models of Evolutionary Algorithms that are extensively used for evolving systems and their optimization. All these differ in their methodology of problem solving. We first present Differential Evolution that uses the differences between individuals for their optimization. The chapter would then present Artificial Immune Systems that are an analogy from the natural immunity system prevalent to fight diseases. Here we discuss the self and non-self methodology of classification. Then we present Co-evolution where the different individuals of the population pool help each other for the evolution. The other topics of discussion include Cultural Algorithm where the evolution is biased by a culture or belief space. We also discuss about Cellular Automata in this chapter.
Anupam Shukla, Ritu Tiwari, Rahul Kala

Hybrid Computing

Frontmatter
Evolutionary Neural and Fuzzy Systems
Abstract
Artificial Neural Networks are valuable tools for problem of machine learning and problem solving. A major problem in the use of neural networks is that the architecture needs to be fixed. Further the training algorithm is needed to fix the various parameters. The training algorithm may many times give a sub-optimal performance by getting trapped in local minima. In this chapter we make use of the optimization powers of the evolutionary algorithms and use them for the construction of the neural network. We would first present the application of evolutionary algorithms in setting the weights and biases of the neural networks. We later make use of evolutionary algorithms for fixing the architecture as well along with the weights and biases. Here we would describe a connectionist approach where the optimal connections evolve. We would also describe an incremental evolution technique in the same problem. We then make use of Grammatical Evolution for evolving the neural networks. At the end we give a similar treatment to the Fuzzy Systems as well.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Modular Neural Networks
Abstract
Modular Neural Networks are use of a number of Neural Networks for problem solving. Here the various neural networks behave as modules to solve a part of the problem. The entire task of division of problem into the various modules as well as the integration of the responses of the modules to generate the final output of the system is done by an integrator. In this chapter we first look at the various Modular Neural Network models. Here we would mainly study two major models. The first model would cluster the entire input space with each module responsible for some part of it. The other model would make different neural networks work over the same problem. Here we would be using a response integration technique for figuring out the final output of the system. The other part of the chapter would present Evolutionary Modular Neural Networks. We would first present a simple genetic approach and then a co-evolutionary approach for this evolution of the entire Modular Neural Network.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Chapter 15 Hybridizing Neural and Fuzzy Systems
Abstract
The neural networks are excellent means of learning where training algorithms may be used for the tuning of the various parameters of the neural network. The fuzzy systems are extensively used for their fuzzy approach to problem modeling and solving. In this chapter we would present how the problem modeling capabilities of the fuzzy systems combines with the learning ability of the neural networks to create the Adaptive Neuro Fuzzy Inference Systems. We later see how these systems may be evolved using an evolutionary approach to make evolutionary neuro fuzzy systems. The other part of the chapter would focus upon the mechanism of fuzzy neural networks. These are neural networks that take fuzzy inputs and generate fuzzy outputs. Here we would transform the various neural computations into fuzzy arithmetic for problem solving. The neural networks are many times regarded as black boxes. We hence need specialized mechanisms to extract out rules from these networks for understanding and implementation. This would be discussed as the last part of the chapter.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Parallel and Hierarchical Evolutionary Algorithms
Abstract
Evolutionary Algorithms are good optimizing agents that use the evolutionary concepts to evolve systems. The complex fitness landscapes are a problem that restricts the performance of these algorithms. Besides specifying the correct parameter is always very important. In this chapter we study the means by which we can execute a number of evolutionary algorithms in a variety of models to give an overall optimal performance to the entire network. We first discuss the Island Model Algorithm where the individuals are evolved in multiple islands with regular interaction. We then study a related algorithm called the Hierarchical Fair Competition algorithm. Here we would separate the various individuals into classes as per their fitness values. Each of these would have a separate evolution. We would then discuss the Nested Evolutionary Strategies where we fix the parameters of an Evolutionary Strategy by some technique. Towards the end we discuss the Hybrid Genetic Algorithm Particle Swarm Optimization. This algorithm generates individuals by twin usage of Genetic Algorithm and Particle Swarm Optimization.
Anupam Shukla, Ritu Tiwari, Rahul Kala

Hybrid Intelligent Systems

Frontmatter
Fusion Methods in Biometrics
Abstract
The use of speech as a biometric gives limited accuracy in the problems of speaker recognition and verification. The need of better recognition scores has resulted in the fusion of speech with other biometric modalities. This chapter discusses the fusion of speech with face which gives a high recognition score at the same time making the system convenient to be used for the user. We discuss three distinct ways to carry out this fusion. The first method is by directly mixing the attributes. This method has problems of excessive dimensionality of the resultant system. Hence many attributes from both modalities need to be deleted. The other method we discuss is the application of modular neural networks with division of attributes. In this technique the various attributes are divided between the various modules. The results are combined by an integrator. The last method is the use of clustering based division of input space by the system.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Multimodal Biometric Systems
Abstract
The uni-modal biometric systems making use of a single biometric modality have a limited performance that restricts their applicability in real life scenarios. The multimodal biometric systems make use of two or more modalities that together achieve much higher performances. In this chapter we discuss the means to fuse three modalities to make a more robust system. We first discuss the fusion of speech, lip, and face. This system uses Hidden Markov Models for the classification and an integration technique called as late integration for decision making from the three modalities. We then discuss the fusion of face, speech and fingerprint. Here each of the individual biometric modalities would make use of modular neural network which would then be combined using a fuzzy integration technique. The last model we discuss would carry the fusion of fingerprint, face and hand geometry. This system uses a variety of fusion techniques including a sum rule, linear discriminant function and decision trees.
Anupam Shukla, Ritu Tiwari, Rahul Kala

Other Supplementary Topics

Frontmatter
Adaptive Systems
Abstract
The task of adjusting the values of the various parameters in the systems, especially the evolutionary systems, leads to a lot of sub-optimality. Besides the ideal parameter values may be different in different contexts and different stages of the algorithm. This necessitates the construction of adaptive systems that can tune their parameters themselves, leading to close to ideal performance. This chapter explores the various types of adaptive systems. The chapter first presents the types of adaptive systems based on their dynamism. Here we discuss the static, deterministic, adaptive and self-adaptive systems. We then discuss the level of adaptation in these systems. Here we present the environment, individual, population and component level adaptation. We present numerous examples of adaptive systems of various kinds.
Anupam Shukla, Ritu Tiwari, Rahul Kala
A Taxonomy of Models
Abstract
The paradigm of hybrid computing deals with the fusion of numerous computing methods, where each method contributes towards the removal of the limitations of the other methods. In this manner we try to magnify the advantages of all methods and remove the limitations of each of the methods. The hybridization deals with the fusion of Artificial Neural Networks, Evolutionary Algorithms, Fuzzy Logic along with heuristics. The large number of models of all of these present a large set of options using which the various methods may be fused. The application areas present further options where each of the models may be customized as per the problem requirements. The same is true with adaptive computing as well, where different adaptation techniques or adaptive algorithms may be used for a better system performance. Complexity largely limits the millions of possibilities of creating different types of hybrid and adaptive systems, which is further emphasized by a limited computation and data availability. This chapter gives a broad overview of the systems presenting the immense varieties in which the systems can be engineered. The basic question that hence arises is how to make a judicious choice amongst these immense options.
Anupam Shukla, Ritu Tiwari, Rahul Kala
A Programmer’s Approach
Abstract
The entire text in this book talks a lot about various kinds of intelligent systems. One of the associated problems with hybrid and adaptive intelligent systems is that these may be complex to code and implement. The lack of implementation may many times lead to a majority of people unable to execute and appreciate much of whatever we present in various chapters. It would hence be wise to invest a few pages speaking about the systems from a programmer’s point of view. This chapter gives some of the implementation details with code snippets that may assist practitioners in engineering these systems. We first have a look at problem solving with neural systems. We would then touch upon the modular neural networks. This would be followed by a brief discussion over the evolutionary neural networks.
Anupam Shukla, Ritu Tiwari, Rahul Kala
Metadaten
Titel
Towards Hybrid and Adaptive Computing
verfasst von
Anupam Shukla
Ritu Tiwari
Rahul Kala
Copyright-Jahr
2010
Verlag
Springer Berlin Heidelberg
Electronic ISBN
978-3-642-14344-1
Print ISBN
978-3-642-14343-4
DOI
https://doi.org/10.1007/978-3-642-14344-1