Dissertations / Theses: 'Genetic algorithm based learning algorithm (GABL)'

1

El-Nainay, Mustafa Y. "Island Genetic Algorithm-based Cognitive Networks." Diss., Virginia Tech, 2009. http://hdl.handle.net/10919/28297.

Full text

Abstract:

The heterogeneity and complexity of modern communication networks demands coupling network nodes with intelligence to perceive and adapt to different network conditions autonomously. Cognitive Networking is an emerging networking research area that aims to achieve this goal by applying distributed reasoning and learning across the protocol stack and throughout the network. Various cognitive node and cognitive network architectures with different levels of maturity have been proposed in the literature. All of them adopt the idea of coupling network devices with sensors to sense network conditions, artificial intelligence algorithms to solve problems, and a reconfigurable platform to apply solutions. However, little further research has investigated suitable reasoning and learning algorithms. In this dissertation, we take cognitive network research a step further by investigating the reasoning component of cognitive networks. In a deviation from previous suggestions, we suggest the use of a single flexible distributed reasoning algorithm for cognitive networks. We first propose an architecture for a cognitive node in a cognitive network that is general enough to apply to future networking challenges. We then introduce and justify our choice of the island genetic algorithm (iGA) as the distributed reasoning algorithm. Having introduced our cognitive node architecture, we then focus on the applicability of the island genetic algorithm as a single reasoning algorithm for cognitive networks. Our approach is to apply the island genetic algorithm to different single and cross layer communication and networking problems and to evaluate its performance through simulation. A proof of concept cognitive network is implemented to understand the implementation challenges and assess the island genetic algorithm performance in a real network environment. We apply the island genetic algorithm to three problems: channel allocation, joint power and channel allocation, and flow routing. The channel allocation problem is a major challenge for dynamic spectrum access which, in turn, has been the focal application for cognitive radios and cognitive networks. The other problems are examples of hard cross layer problems. We first apply the standard island genetic algorithm to a channel allocation problem formulated for the dynamic spectrum cognitive network environment. We also describe the details for implementing a cognitive network prototype using the universal software radio peripheral integrated with our extended implementation of the GNU radio software package and our island genetic algorithm implementation for the dynamic spectrum channel allocation problem. We then develop a localized variation of the island genetic algorithm, denoted LiGA, that allows the standard island genetic algorithm to scale and apply it to the joint power and channel allocation problem. In this context, we also investigate the importance of power control for cognitive networks and study the effect of non-cooperative behavior on the performance of the LiGA. The localized variation of the island genetic algorithm, LiGA, is powerful in solving node-centric problems and problems that requires only limited knowledge about network status. However, not every communication and networking problems can be solved efficiently in localized fashion. Thus, we propose a generalized version of the LiGA, namely the K-hop island genetic algorithm, as our final distributed reasoning algorithm proposal for cognitive networks. The K-hop island genetic algorithm is a promising algorithm to solve a large class of communication and networking problems with controllable cooperation and migration scope that allows for a tradeoff between performance and cost. We apply it to a flow routing problem that includes both power control and channel allocation. For all problems simulation results are provided to quantify the performance of the island genetic algorithm variation. In most cases, simulation and experimental results reveal promising performance for the island genetic algorithm. We conclude our work with a discussion of the shortcomings of island genetic algorithms without guidance from a learning mechanism and propose the incorporation of two learning processes into the cognitive node architecture to solve slow convergence and manual configuration problems. We suggest the cultural algorithm framework and reinforcement learning techniques as candidate leaning techniques for implementing the learning processes. However, further investigation and implementation is left as future work.
Ph. D.

APA, Harvard, Vancouver, ISO, and other styles

2

Tamaddoni, Nezhad Alireza. "Logic-based machine learning using a bounded hypothesis space : the lattice structure, refinement operators and a genetic algorithm approach." Thesis, Imperial College London, 2013. http://hdl.handle.net/10044/1/29849.

Full text

Abstract:

Rich representation inherited from computational logic makes logic-based machine learning a competent method for application domains involving relational background knowledge and structured data. There is however a trade-off between the expressive power of the representation and the computational costs. Inductive Logic Programming (ILP) systems employ different kind of biases and heuristics to cope with the complexity of the search, which otherwise is intractable. Searching the hypothesis space bounded below by a bottom clause is the basis of several state-of-the-art ILP systems (e.g. Progol and Aleph). However, the structure of the search space and the properties of the refinement operators for theses systems have not been previously characterised. The contributions of this thesis can be summarised as follows: (i) characterising the properties, structure and morphisms of bounded subsumption lattice (ii) analysis of bounded refinement operators and stochastic refinement and (iii) implementation and empirical evaluation of stochastic search algorithms and in particular a Genetic Algorithm (GA) approach for bounded subsumption. In this thesis we introduce the concept of bounded subsumption and study the lattice and cover structure of bounded subsumption. We show the morphisms between the lattice of bounded subsumption, an atomic lattice and the lattice of partitions. We also show that ideal refinement operators exist for bounded subsumption and that, by contrast with general subsumption, efficient least and minimal generalisation operators can be designed for bounded subsumption. In this thesis we also show how refinement operators can be adapted for a stochastic search and give an analysis of refinement operators within the framework of stochastic refinement search. We also discuss genetic search for learning first-order clauses and describe a framework for genetic and stochastic refinement search for bounded subsumption. on. Finally, ILP algorithms and implementations which are based on this framework are described and evaluated.

APA, Harvard, Vancouver, ISO, and other styles

3

Suleiman, Iyad. "Integrating data mining and social network techniques into the development of a Web-based adaptive play-based assessment tool for school readiness." Thesis, University of Bradford, 2013. http://hdl.handle.net/10454/7293.

Full text

Abstract:

A major challenge that faces most families is effectively anticipating how ready to start school a given child is. Traditional tests are not very effective as they depend on the skills of the expert conducting the test. It is argued that automated tools are more attractive especially when they are extended with games capabilities that would be the most attractive for the children to be seriously involved in the test. The first part of this thesis reviews the school readiness approaches applied in various countries. This motivated the development of the sophisticated system described in the thesis. Extensive research was conducted to enrich the system with features that consider machine learning and social network aspects. A modified genetic algorithm was integrated into a web-based stealth assessment tool for school readiness. The research goal is to create a web-based stealth assessment tool that can learn the user's skills and adjust the assessment tests accordingly. The user plays various sessions from various games, while the Genetic Algorithm (GA) selects the upcoming session or group of sessions to be presented to the user according to his/her skills and status. The modified GA and the learning procedure were described. A penalizing system and a fitness heuristic for best choice selection were integrated into the GA. Two methods for learning were presented, namely a memory system and a no-memory system. Several methods were presented for the improvement of the speed of learning. In addition, learning mechanisms were introduced in the social network aspect to address further usage of stealth assessment automation. The effect of the relatives and friends on the readiness of the child was studied by investigating the social communities to which the child belongs and how the trend in these communities will reflect on to the child under investigation. The plan is to develop this framework further by incorporating more information related to social network construction and analysis. Also, it is planned to turn the framework into a self adaptive one by utilizing the feedback from the usage patterns to learn and adjust the evaluation process accordingly.

APA, Harvard, Vancouver, ISO, and other styles

4

Le, Bin. "Building a Cognitive Radio: From Architecture Definition to Prototype Implementation." Diss., Virginia Tech, 2007. http://hdl.handle.net/10919/28320.

Full text

Abstract:

Cognitive radio (CR) technology introduces a revolutionary wireless communication mechanism in terminals and network segments, so that they are able to learn their environment and adapt intelligently to the most appropriate way of providing the service for the user's exact need. By supporting multi-band, mode-mode cognitive applications, the cognitive radio addresses an interactive way of managing the spectrum that harmonizes technology, market and regulation. This dissertation gives a complete story of building a cognitive radio. It goes through concept clarification, architecture definition, functional block building, system integration, and finally to the implementation of a fully-functional cognitive radio node prototype that can be directly packaged for application use. This dissertation starts with a comprehensive review of CR research from its origin to today. Several fundamental research issues are then addressed to let the reader know what makes CR a challenging and interesting research area. Then the CR system solution is introduced with the details of its hierarchical functional architecture called the Egg Model, modular software system called the cognitive engine, and the kernel machine learning mechanism called the cognition cycle. Next, this dissertation discusses the design of specific functional building blocks which incorporate environment awareness, solution making, and adaptation. These building blocks are designed to focus on the radio domain that mainly concerns the radio environment and the radio platform. Awareness of the radio environment is achieved by extracting the key environmental features and applying statistical pattern recognition methods including artificial neural networks and k-nearest neighbor clustering. Solutions for the radio behavior are made according to the recognized environment and the previous knowledge through case based reasoning, and further adapted or optimized through genetic algorithm solution search. New experiences are gained through the practice of the new solution, and thus the CR's knowledge evolves for future use; therefore, the CR's performance continues improving with this reinforcement learning approach. To deploy the solved solution in terms of the radio's parameters, a platform independent radio interface is designed. With this general radio interface, the algorithms in the cognitive engine software system can be applied to various radio hardware platforms. To support and verify designed cognitive algorithms and cognitive functionalities, a complete reconfigurable SDR platform, called the CWT2 waveform framework, is designed in this dissertation. In this waveform framework, a hierarchical configuration and control system is constructed to support flexible, real-time waveform reconfigurability. Integrating all the building blocks described above allows a complete CR node system. Based on this general CR node structure, a fully-functional Public Safety Cognitive Radio (PSCR) node is prototyped to provide the universal interoperability for public safety communications. Although the complete PSCR node software system has been packaged to an official release including installation guide and user/developer manuals, the process of building a cognitive radio from concept to a functional prototype is not the end of the CR research; on-going and future research issues are addressed in the last chapter of the dissertation.
Ph. D.

APA, Harvard, Vancouver, ISO, and other styles

5

Almejalli, Khaled A. "Intelligent Real-Time Decision Support Systems for Road Traffic Management. Multi-agent based Fuzzy Neural Networks with a GA learning approach in managing control actions of road traffic centres." Thesis, University of Bradford, 2010. http://hdl.handle.net/10454/4264.

Full text

Abstract:

The selection of the most appropriate traffic control actions to solve non-recurrent traffic congestion is a complex task which requires significant expert knowledge and experience. In this thesis we develop and investigate the application of an intelligent traffic control decision support system for road traffic management to assist the human operator to identify the most suitable control actions in order to deal with non-recurrent and non-predictable traffic congestion in a real-time situation. Our intelligent system employs a Fuzzy Neural Networks (FNN) Tool that combines the capabilities of fuzzy reasoning in measuring imprecise and dynamic factors and the capabilities of neural networks in terms of learning processes. In this work we present an effective learning approach with regard to the FNN-Tool, which consists of three stages: initializing the membership functions of both input and output variables by determining their centres and widths using self-organizing algorithms; employing an evolutionary Genetic Algorithm (GA) based learning method to identify the fuzzy rules; tune the derived structure and parameters using the back-propagation learning algorithm. We evaluate experimentally the performance and the prediction capability of this three-stage learning approach using well-known benchmark examples. Experimental results demonstrate the ability of the learning approach to identify all relevant fuzzy rules from the training data. A comparative analysis shows that the proposed learning approach has a higher degree of predictive capability than existing models. We also address the scalability issue of our intelligent traffic control decision support system by using a multi-agent based approach. The large network is divided into sub-networks, each of which has its own associated agent. Finally, our intelligent traffic control decision support system is applied to a number of road traffic case studies using the traffic network in Riyadh, in Saudi Arabia. The results obtained are promising and show that our intelligent traffic control decision support system can provide an effective support for real-time traffic control.

APA, Harvard, Vancouver, ISO, and other styles

6

Dam, Hai Huong Information Technology &amp Electrical Engineering Australian Defence Force Academy UNSW. "A scalable evolutionary learning classifier system for knowledge discovery in stream data mining." Awarded by:University of New South Wales - Australian Defence Force Academy, 2008. http://handle.unsw.edu.au/1959.4/38865.

Full text

Abstract:

Data mining (DM) is the process of finding patterns and relationships in databases. The breakthrough in computer technologies triggered a massive growth in data collected and maintained by organisations. In many applications, these data arrive continuously in large volumes as a sequence of instances known as a data stream. Mining these data is known as stream data mining. Due to the large amount of data arriving in a data stream, each record is normally expected to be processed only once. Moreover, this process can be carried out on different sites in the organisation simultaneously making the problem distributed in nature. Distributed stream data mining poses many challenges to the data mining community including scalability and coping with changes in the underlying concept over time. In this thesis, the author hypothesizes that learning classifier systems (LCSs) - a class of classification algorithms - have the potential to work efficiently in distributed stream data mining. LCSs are an incremental learner, and being evolutionary based they are inherently adaptive. However, they suffer from two main drawbacks that hinder their use as fast data mining algorithms. First, they require a large population size, which slows down the processing of arriving instances. Second, they require a large number of parameter settings, some of them are very sensitive to the nature of the learning problem. As a result, it becomes difficult to choose a right setup for totally unknown problems. The aim of this thesis is to attack these two problems in LCS, with a specific focus on UCS - a supervised evolutionary learning classifier system. UCS is chosen as it has been tested extensively on classification tasks and it is the supervised version of XCS, a state of the art LCS. In this thesis, the architectural design for a distributed stream data mining system will be first introduced. The problems that UCS should face in a distributed data stream task are confirmed through a large number of experiments with UCS and the proposed architectural design. To overcome the problem of large population sizes, the idea of using a Neural Network to represent the action in UCS is proposed. This new system - called NLCS { was validated experimentally using a small fixed population size and has shown a large reduction in the population size needed to learn the underlying concept in the data. An adaptive version of NLCS called ANCS is then introduced. The adaptive version dynamically controls the population size of NLCS. A comprehensive analysis of the behaviour of ANCS revealed interesting patterns in the behaviour of the parameters, which motivated an ensemble version of the algorithm with 9 nodes, each using a different parameter setting. In total they cover all patterns of behaviour noticed in the system. A voting gate is used for the ensemble. The resultant ensemble does not require any parameter setting, and showed better performance on all datasets tested. The thesis concludes with testing the ANCS system in the architectural design for distributed environments proposed earlier. The contributions of the thesis are: (1) reducing the UCS population size by an order of magnitude using a neural representation; (2) introducing a mechanism for adapting the population size; (3) proposing an ensemble method that does not require parameter setting; and primarily (4) showing that the proposed LCS can work efficiently for distributed stream data mining tasks.

APA, Harvard, Vancouver, ISO, and other styles

7

Castro, Neto Henrique de. "Uma nova abordagem de aprendizagem de máquina combinando elicitação automática de casos, aprendizagem por reforço e mineração de padrões sequenciais para agentes jogadores de damas." Universidade Federal de Uberlândia, 2016. https://repositorio.ufu.br/handle/123456789/18143.

Full text

Abstract:

Fundação de Amparo a Pesquisa do Estado de Minas Gerais
Agentes que operam em ambientes onde as tomadas de decisão precisam levar em conta, além do ambiente, a atuação minimizadora de um oponente (tal como nos jogos), é fundamental que o agente seja dotado da habilidade de, progressivamente, traçar um perĄl de seu adversário que o auxilie em seu processo de seleção de ações apropriadas. Entretanto, seria improdutivo construir um agente com um sistema de tomada de decisão baseado apenas na elaboração desse perĄl, pois isso impediria o agente de ter uma Şidentidade própriaŤ, o que o deixaria a mercê de seu adversário. Nesta direção, este trabalho propõe um sistema automático jogador de Damas híbrido, chamado ACE-RL-Checkers, dotado de um mecanismo dinâmico de tomada de decisões que se adapta ao perĄl de seu oponente no decorrer de um jogo. Em tal sistema, o processo de seleção de ações (movimentos) é conduzido por uma composição de Rede Neural de Perceptron Multicamadas e biblioteca de casos. No caso, a Rede Neural representa a ŞidentidadeŤ do agente, ou seja, é um módulo tomador de decisões estático já treinado e que faz uso da técnica de Aprendizagem por Reforço TD( ). Por outro lado, a biblioteca de casos representa o módulo tomador de decisões dinâmico do agente que é gerada pela técnica de Elicitação Automática de Casos (um tipo particular de Raciocínio Baseado em Casos). Essa técnica possui um comportamento exploratório pseudo-aleatório que faz com que a tomada de decisão dinâmica do agente seja guiada, ora pelo perĄl de jogo do adversário, ora aleatoriamente. Contudo, ao conceber tal arquitetura, é necessário evitar o seguinte problema: devido às características inerentes à técnica de Elicitação Automática de Casos, nas fases iniciais do jogo Ű em que a quantidade de casos disponíveis na biblioteca é extremamente baixa em função do exíguo conhecimento do perĄl do adversário Ű a frequência de tomadas de decisão aleatórias seria muito elevada, o que comprometeria o desempenho do agente. Para atacar tal problema, este trabalho também propõe incorporar à arquitetura do ACE-RLCheckers um terceiro módulo, composto por uma base de regras de experiência extraída a partir de jogos de especialistas humanos, utilizando uma técnica de Mineração de Padrões Sequenciais. O objetivo de utilizar tal base é reĄnar e acelerar a adaptação do agente ao perĄl de seu adversário nas fases iniciais dos confrontos entre eles. Resultados experimentais conduzidos em torneio envolvendo ACE-RL-Checkers e outros agentes correlacionados com este trabalho, conĄrmam a superioridade da arquitetura dinâmica aqui proposta.
ake into account, in addition to the environment, the minimizing action of an opponent (such as in games), it is fundamental that the agent has the ability to progressively trace a proĄle of its adversary that aids it in the process of selecting appropriate actions. However, it would be unsuitable to construct an agent with a decision-making system based on only the elaboration of this proĄle, as this would prevent the agent from having its Şown identityŤ, which would leave it at the mercy of its opponent. Following this direction, this work proposes an automatic hybrid Checkers player, called ACE-RL-Checkers, equipped with a dynamic decision-making mechanism, which adapts to the proĄle of its opponent over the course of the game. In such a system, the action selection process (moves) is conducted through a composition of Multi-Layer Perceptron Neural Network and case library. In the case, Neural Network represents the ŞidentityŤ of the agent, i.e., it is an already trained static decision-making module and makes use of the Reinforcement Learning TD( ) techniques. On the other hand, the case library represents the dynamic decision-making module of the agent, which is generated by the Automatic Case Elicitation technique (a particular type of Case-Based Reasoning). This technique has a pseudo-random exploratory behavior, which makes the dynamic decision-making on the part of the agent to be directed, either by the game proĄle of the opponent or randomly. However, when devising such an architecture, it is necessary to avoid the following problem: due to the inherent characteristics of the Automatic Case Elicitation technique, in the game initial phases, in which the quantity of available cases in the library is extremely low due to low knowledge content concerning the proĄle of the adversary, the decisionmaking frequency for random decisions is extremely high, which would be detrimental to the performance of the agent. In order to attack this problem, this work also proposes to incorporate onto the ACE-RL-Checkers architecture a third module composed of a base of experience rules, extracted from games played by human experts, using a Sequential Pattern Mining technique. The objective behind using such a base is to reĄne and accelerate the adaptation of the agent to the proĄle of its opponent in the initial phases of their confrontations. Experimental results conducted in tournaments involving ACE-RL-Checkers and other agents correlated with this work, conĄrm the superiority of the dynamic architecture proposed herein.
Tese (Doutorado)

APA, Harvard, Vancouver, ISO, and other styles

8

(20390), Baolin Wu. "Fuzzy modelling and identification with genetic algorithms based learning." Thesis, 1996. https://figshare.com/articles/thesis/Fuzzy_modelling_and_identification_with_genetic_algorithms_based_learning/21345057.

Full text

Abstract:

Modelling is an essential step towards a solution to complex system problems. Traditional mathematical methods are inadequate in describing the complex systems when the complexity increases. Fuzzy logic has provided an alternative way in dealing with complexity in real world.

This thesis looks at a practical approach for complex system modelling using fuzzy logic. This approach is usually called fuzzy modelling. The main aim of this thesis is to explore the capabilities of fuzzy logic in complex system modelling using available data. The fuzzy model concerned is the Sugeno-Takage-Kang model (TSK model). A genetic algorithm based learning algorithm (GABL) is proposed for fuzzy modelling. It basically contains four blocks, namely the partition, GA, tuning and termination blocks. The functioning of each block is described and the proposed algorithm is tested using a number of examples from different applications such as function approximation and processing control.

APA, Harvard, Vancouver, ISO, and other styles

9

Weng, Kuei-Sung, and 翁桂松. "Fuzzy Modeling Based on Genetic Ellipsoid Learning Algorithm." Thesis, 2002. http://ndltd.ncl.edu.tw/handle/74814313454610298473.

Full text

Abstract:

碩士
國立臺北科技大學
機電整合研究所
90
The theme of this thesis is to apply Genetic Algorithm (GA) and Gustafson-Kessel (G-K) Algorithm to the fuzzy modeling. A method called Genetic Ellipsoid Learning Algorithm (GELA) is proposed to learn the decision regions for pattern recognition and adaptive fuzzy modeling in this thesis. 1.First topics, a learning method based on fuzzy clustering and adaptively tuned hyperellipsoids is proposed to learn the decision regions for pattern recognition. The Gustafson- Kessel (G-K) algorithm for fuzzy clustering is modified in such a way that the Genetic Algorithm is applied to dynamically learn the volumes of hyperellipsoids in G-K algorithm. The decision regions are accurately learned by the proposed method in this paper so that on one hand, misclassification errors are minimized; on the other hand, the range of learned decision regions are not too wide to reduce the accuracy of pattern recognition. 2.Second topics, a method called Genetic Ellipsoid Learning Algorithm (GELA) is proposed for adaptive fuzzy modeling integrating Genetic Algorithm (GA) and Gustafson-Kessel (G-K) Algorithm. Since G-K algorithm is able to efficiently cover data points with multiple ellipsoids, GA is applied to estimate volume of each ellipsoid. Based on the volume learned by GA as well as input/output data points, G-K algorithm will then estimate the parameters of each ellipsoid. As input/output data points are clustered by multiple ellipsoids, a second GA is proposed to fine-tune the parameters of each ellipsoid for fuzzy modeling.

APA, Harvard, Vancouver, ISO, and other styles

10

Tzou, Tsung-Fei, and 鄒璁飛. "The Reinforcement Learning Behavior Unit Weights Searching based on Genetic Algorithm." Thesis, 2007. http://ndltd.ncl.edu.tw/handle/67380234386530334498.

Full text

Abstract:

碩士
國立中正大學
電機工程所
95
This thesis proposes a scheme based on Stochastic Searching Network and (GA) Genetic Algorithm, and we use Reinforcement Learning method for action network weights searching problem. The SGRL learning scheme is a hybrid Genetic Algorithm, which integrates the Stochastic Searching Network and the Genetic Algorithm to fulfill the Reinforcement Learning action network weights searching task. Structurally, the SGRL learning system is composed of two integrated feed-forward networks. One neural network acts as a critic network for helping the learning of the other network, the action network, which determines the outputs (actions) of the SGRL learning system, where the action network is a normal neural network. Using the TD (Temporal Difference) prediction method, the critic network can predict the external reinforcement signal and provide a more informative internal reinforcement signal to the action network. The action network uses the GA and according to the plant dynamic reference　 model to adapt itself according to the internal reinforcement signal. The key concept of the SGRL learning scheme is to formulate the internal reinforcement signal contributed by the reference plant model as the fitness function for the GA. Computer simulations on controlling of the Acrobot (i.e. possessing fewer actuators than degrees of freedom) system and mountain-car system have been conducted to illustrate the performance and applicability of the proposed learning controller scheme.

APA, Harvard, Vancouver, ISO, and other styles

11

Pan, Pang-Chi, and 潘邦積. "A Genetic Algorithm based on Reinforcement Learning for the Traveling Salesman Problem." Thesis, 2008. http://ndltd.ncl.edu.tw/handle/66211216522086816285.

Full text

Abstract:

碩士
國立高雄第一科技大學
電腦與通訊工程所
96
In this thesis, a novel genetic-algorithm-based approach for finding approximate solutions to a general version of the well-known Traveling Salesman Problem (TSP) is presented. On the basis of the concepts of reinforcement learning, several sophisticated operators are introduced to explicitly balance the exploration and exploitation abilities of the proposed algorithm. The performance of the proposed algorithm is evaluated by comparing it against existing genetic-algorithm-based techniques in terms of overall schedule length for a set of problem instances obtained form the traveling salesman problem library (TSPLIB). Experimental results indicate that the algorithm proposed here outperforms the conventional approaches.

APA, Harvard, Vancouver, ISO, and other styles

12

Hsieh, Su-Hwang, and 謝書桓. "Genetic Algorithm Based Fuzzy ID3 Method for Data Learning with Mixed-Mode Attributes." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/xgr7ts.

Full text

Abstract:

碩士
國立交通大學
電機與控制工程系所
92
Many learning approaches to knowledge acquisition have been promisingly developed recently. A popular and efficient method for decision tree induction from discrete data is ID3 algorithm. However, most knowledge associated with human’s thinking and perception has some imprecision and uncertainty. For the purpose of handling imprecise and uncertain knowledge, the decision tree induction has been improved so that it is suitable for the fuzzy case. Several fuzzy ID3 schemes were proposed, but they can only deal with continuous data and are often criticized to result in poor learning accuracy. In this thesis, we propose a method to generate a fuzzy decision tree, which can accept continuous, discrete, or mixed-mode data and it is designed based on genetic algorithm. Next, we formulated a pruning method for our algorithm to obtain a more compact rule-base. We have tested our method on ten data sets from the UCI Repository, and the results of a two-fold cross validation are compared to those by C5.0. The experiments show that our method works better in practice. Finally, we analysis a web log-file data set using our fuzzy ID3 method, the rule-base extracted from the fuzzy ID3 decision tree can provide important directions to web master for improve the contents of the website.

APA, Harvard, Vancouver, ISO, and other styles

13

Hui-WenTien and 田惠文. "A Genetic Algorithm-Based Multi-Characteristic Decision-Making Grouping Strategy for Collaborative Learning." Thesis, 2013. http://ndltd.ncl.edu.tw/handle/17488142291821671414.

Full text

Abstract:

碩士
國立成功大學
資訊工程學系碩博士班
101
In collaborative learning, previous researches indicated that well-balanced groups could enhance students’ learning performance. However, to construct well-balanced groups for large number of students with multiple characteristics will cost considerable efforts and time to instructors. Hence, how to automatically construct well-balanced learning groups has been an important issue in collaborative learning. This thesis proposes a new grouping strategy based on genetic algorithm (GA) with Technique for Order Performance by Similarity to Ideal Solution (TOPSIS) to assist instructors in constructing inter-homogeneous and intra-heterogeneous collaborative learning groups. The grouping strategy proposed solves the multi-characteristic grouping problem, where the characteristics are in the same direction or in different directions. Further, the constituent members of groups can be balanced to achieve the inter-groups homogeneity and intra-groups heterogeneity. Meanwhile, several datasets with different problem sizes including an actual case were employed as experimental materials to conduct a series of experiments. Experimental results demonstrated that the proposed grouping method outperforms the random and genetic algorithm methods in terms of solution quality, executing time, search speed and stability. To sum, the proposed grouping method is more effective, efficient, robust, and has better grouping result and quality.

APA, Harvard, Vancouver, ISO, and other styles

14

Shiau, Jing-Wen, and 蕭景文. "A Genetic Algorithm Learning Based CMAC with Gaussian Basis Function and Its Application in Function Learning." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/54869360266144201231.

Full text

Abstract:

碩士
大同大學
電機工程學系(所)
93
Cerebellar Model Arithmetic Controller (CMAC) is one of neural networks and its advantage is fast learning property, good generalization capability, and ease of implementation by hardware. It is, however, difficult to decide various parameters of CMAC in advance. Genetic Algorithm (GA) is one of Evolutionary Algorithms (EAs), and is efficient in local search. Employing genetic algorithms on the design and training of CMAC allows the CMAC parameters to be easily optimized. CMAC can be viewed as a radial basis function (RBF) network. The conventional CMAC uses a local constant basis function (also called rectangle function) to model the hypercube structure. A disadvantage is that its output is always constant within each quantized state and the derivative information of input and output variables cannot be preserved. If the local constant basis functions are replaced by non-constant differentiable basis functions, the derivative information will be able to be stored into the structure as well. Therefore, we use Gaussian basis function (GBF) to improve the accuracy of GA-CMAC. In the experimental results, the GA-CMAC with GBF is performed to demonstrate the improvement of accuracy in modeling.

APA, Harvard, Vancouver, ISO, and other styles

15

Haque, Mohammad Nazmul. "Genetic algorithm-based ensemble methods for large-scale biological data classification." Thesis, 2017. http://hdl.handle.net/1959.13/1335393.

Full text

Abstract:

Research Doctorate - Doctor of Philosophy (PhD)
We study the search for the best ensemble combinations from the wide variety of heterogeneous base classifiers. The number of possible ways to create the ensemble with a large number of base classifiers is exponential to the base classifiers pool size. To search for the best combinations from that wide search space is not suitable for exhaustive search because of it's exponential growth with the ensemble size. Hence, we employed a genetic algorithm to find the best ensemble combinations from a pool of heterogeneous base classifiers. The classification decisions of base classifiers are combined using the popular majority vote approach. We used random sub-sampling for balancing the class distributions in the class-imbalanced datasets. The empirical result on benchmarking and real-world datasets apparently outperformed the performances of base classifiers and other state-of-the-art ensemble methods. Afterwards, we evaluated the performance of an ensemble of classifiers combination search in a weighted voting approach using the differential evolution (DE) algorithm to find if employing weights could increase the generalisation performances of ensembles. The weights optimised by DE also outperformed both of the base classifiers and other ensembles for benchmarking and real-world biological datasets. Finally, we extend the majority voting-based ensemble of classifiers combination search with multi-objective settings. The search space is spread over the all possible ensemble combinations created with 29 heterogeneous base classifiers and the selection of feature subset from six feature selection methods as wrapper approach. The optimisation of two objectives, the maximisation of training MCC scores and maximisation of the diversity among base classifiers, with NSGA-II, a popular multi-objective genetic algorithm, is used for simultaneously finding the best feature set and the ensemble combinations. We analyse the Pareto front of solutions obtained by NSGA-II for their generalisation performances. Datasets taken from UCI machine learning repository and NIPS2003 feature selection challenges have been used to investigate the performance of proposed method. The experimental outcomes suggest that the proposed multiobjective-based NSGA-II found the better feature set and the best ensemble combination that produces better generalisation performances in compared to other ensemble of classifiers methods.

APA, Harvard, Vancouver, ISO, and other styles

16

Stewart, IAN. "A Modified Genetic Algorithm and Switch-Based Neural Network Model Applied to Misuse-Based Intrusion Detection." Thesis, 2009. http://hdl.handle.net/1974/1720.

Full text

Abstract:

As our reliance on the Internet continues to grow, the need for secure, reliable networks also increases. Using a modified genetic algorithm and a switch-based neural network model, this thesis outlines the creation of a powerful intrusion detection system (IDS) capable of detecting network attacks. The new genetic algorithm is tested against traditional and other modified genetic algorithms using common benchmark functions, and is found to produce better results in less time, and with less human interaction. The IDS is tested using the standard benchmark data collection for intrusion detection: the DARPA 98 KDD99 set. Results are found to be comparable to those achieved using ant colony optimization, and superior to those obtained with support vector machines and other genetic algorithms.
Thesis (Master, Computing) -- Queen's University, 2009-03-03 13:28:23.787

APA, Harvard, Vancouver, ISO, and other styles

17

Wang, Nien-Hung, and 王年宏. "Using FP-growth Genetic Algorithm to Construct Neural-Fuzzy Control Systems Based on Reinforcement Learning Scheme." Thesis, 2008. http://ndltd.ncl.edu.tw/handle/68499373612441324994.

Full text

Abstract:

碩士
國立交通大學
電機與控制工程系所
96
Recently, evolutionary algorithms are widely applied in several regions. In the traditionalgenetic algorithm, the chromosome is evolved by using random search to execute crossover and mutation. However, the gene with good performance may be tuned repeatedly and the evolutionary time will be much longer. In this thesis, a genetic algorithm based on data mining is adopted to solve this problem. By using the ability of looking for association data, the gene point associating with the improvement of fitness value can be found. Hence, suitable crossover points and mutation points can be found systematically, and the algorithm can be converged more efficiently. In this thesis, a neural-fuzzy control system using reinforcement learning is constructed. This thesis also uses the technique of data mining to enhance the evolution efficiency of the algorithm. The asymmetric crossover and mutation is proposed to learn the parameters and structure of neural-fuzzy control system. The ball and beam control system and chaos control system are used as examples to test learning ability and controllability of the proposed system. The experimental results show the system is satisfactory.

APA, Harvard, Vancouver, ISO, and other styles

18

Lu, Chi-Fu, and 盧啟富. "Fuzzy Logic Models with Adaptive Learning Rates and Genetic Algorithm for Thermally Based Microelectronic Manufacturing Processes." Thesis, 1996. http://ndltd.ncl.edu.tw/handle/77342210875456884445.

Full text

Abstract:

碩士
國立交通大學
控制工程系
84
This paper presents the improved fuzzy logic models (FLM) to simulate the thermally based microelectronic manufacturing process: the silicon deposition process in a barrel chemical vapor deposition (CVD) reactor. To identify a FLM for a process, there are two major tasks: structure and parameter identifications. In structure identification, the genetic algorithm is used to search for the optimal structure so that the predictive capability of the FLM is increased. In parameter identification, the adaptive learning rate that is based on the sum of square errors between given data and output of the FLM is chosen to increase the convergent speed of the parameters. Several mathematical functions and a CVD process are used to demonstrate the efficiency and accuracy of the improved FLM in comparison with the existing fuzzy models.

APA, Harvard, Vancouver, ISO, and other styles

19

高秀婷. "Application of the Hadoop-based Parallel Genetic Algorithm to the Identification of Students with Learning Disabilies." Thesis, 2011. http://ndltd.ncl.edu.tw/handle/38611900182237267236.

Full text

Abstract:

碩士
國立彰化師範大學
數位內容科技與管理研究所
99
Process in identification of students with Learning Disabilities (LD) may require a lot of time and manpower. In order to reduce the evaluation personnel's workload, researchers use Artificial Neural Network (ANN) to develop an assisted LD students identification system. The results shows that through MPI-based Distributed Parallel Genetic Algorithm (DPGA) with ANN, the operation time and identfication accuracy can both be improved. However, the MPI-based solution may have the communication and data storage issues in the distrubuted environment. In this research, we port the assisted LD students identification system to the open-source Hadoop framework. We resesign the parallel genetic algorithm and map it to the MapReduce program framework, and store corresponding data in the Haddop Data File System. The solution can reduce the programers loading on taking care of the communication and data storage issues in the distrubuted environment. To fit our problem into the Hadoop environment, we design a data file segmentation approach. Our experiment results show that the developed method may be 6.8 times faster that the Hadoop-default segmentation algorithm. In addition, we use HDFS to store the elite chromosome, and as the communication media among the computation nodes. By setting a threshold value to determine whether the elite chromosomes should be kept for the later generation, we may get a higher accuracy in correct identification of students with learning disabilities. By comparing results to previous study using MPI-based solution, we have found that our solution may have edge in the case of larger data set.

APA, Harvard, Vancouver, ISO, and other styles

20

Ke, Yan-Ru, and 柯彥如. "A Personalized e-Course Composition based on Genetic Algorithm with Forcing Legality in an Adaptive Learning System." Thesis, 2011. http://ndltd.ncl.edu.tw/handle/28504335775415228543.

Full text

Abstract:

碩士
國立彰化師範大學
數位學習研究所
99
This paper proposes a personalized e-course composition based on a genetic algorithm with forcing legality in adaptive learning systems, which efficiently and accurately finds appropriate e-learning materials in the database for individual learners. The algorithm not only reduces the search space size and increases search efficiency but also is more explicit in finding the best e-course composition in a legal solution space. The serial experiments indicate that the genetic algorithm with forcing legality regardless of the number of students or the number of materials in the database, to compose a personalized e-course within a limited time is much more efficient and accurate than the method based on the particle swarm optimization proposed by Chu et al. and the improved particle swarm optimization proposed by Dheeban et al. Therefore, the genetic algorithm with forcing legality is able to enhance the quality of personalized e-course compositions in adaptive learning environments.

APA, Harvard, Vancouver, ISO, and other styles

21

ho, ka-king, and 何家敬. "Adaptive Network-Based Fuzzy Inference System for Prediction of Workpiece Surface Roughness in End Milling Using Genetic-Based Learning Algorithm." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/77207990673716195157.

Full text

Abstract:

碩士
國立高雄第一科技大學
機械與自動化工程所
95
In this paper , genetic-based learning algorithm method of adaptive-network based fuzzy inference system(ANFIS) was used to predict the work piece surface roughness. Spindle speed , feed rate and depth of cut are the input variable , and the final output is the surface roughness . The premise part of the ANFIS is applied by Gauss membership function , and the consequent part is applied by the Takagi-Sugeno (TS) fuzzy model to inference. Taguchi-genetic algorithm was used to train the parameters of both the Gauss membership function in the premise part and the function in consequent part. A total of 48 sets of experimental date were used for training. After the training, another 24 sets dates were used to check out how correct the results are. Finally, we compared the prediction accuracy of surface roughness by Genetic-based learning Algorithm method of adaptive-network based fuzzy inference system and the general ANFIS which include triangular and trapezoidal membership function . The comparison indicates that the adoption of genetic-based learning algorithm method get a better performance.

APA, Harvard, Vancouver, ISO, and other styles

22

Liu, Yu-Chen, and 劉祐辰. "Developing the Fuzzy Query and Intelligent Recommendation Mechanism for Smart Kitchen App Based on Machine Learning and Genetic Algorithm." Thesis, 2019. http://ndltd.ncl.edu.tw/handle/zr52ry.

Full text

Abstract:

碩士
國立彰化師範大學
資訊工程學系
108
Nowadays, more people are choosing to cook and eat at home, and the need to look for recipes on websites is increasing. Conventional recipe websites often suffer several problems: the lack of a united website for recipes, the information overload of searching results caused by keywords filtering, and the conditions cannot be considered simultaneously. Furthermore, the lack of personalized services based on personal preference. To alleviate the mentioned problems, we develop an APP, called Smart Kitchen. First, a crawler is developed to capture and analyze recipes data on websites, a reinforcement learning method of CLIPS is applied to classify the ingredients of recipes, and a text classifier called FastText is used to unify the classification of recipes through machine learning. Second, fuzzy theory and constraints satisfaction are combined to implement fuzzy query system, calculate the overall degree of satisfaction based on degrees of importance and satisfaction of each group of query conditions and display query results by descending order. Finally, we implement three intelligent recommendations, including popularity recommendations for operations of all users within recent 30 days, personalized recommendations for personal preference, and collaborative filtering recommendations for users with same favorite. A genetic algorithm is applied to gradually improve the recommendation results by continuous adaptation according to users' feedbacks.

APA, Harvard, Vancouver, ISO, and other styles

23

Luo, Shin-Yi, and 羅欣儀. "A combined genetic algorithm and simulated annealing approach for two-agent scheduling with sum-of-processing-times-based learning effect." Thesis, 2012. http://ndltd.ncl.edu.tw/handle/8a6m32.

Full text

Abstract:

碩士
逢甲大學
統計與精算所
100
This paper considers a scheduling model involving two agents, job ready times, and the sum-of-processing-times-based learning effect. The sum-of-processing-times-based learning effect means that the actual processing time of a job of either agent is a decreasing function of the sum of the processing times of the jobs already scheduled in a given schedule. The goal is to find an optimal schedule that minimizes the total weighted completion time of the first agent, subject to no tardy job for the second agent. We first provide a branch-and-bound method to solve the problem. We then develop an approach that combines genetic algorithm and simulated annealing to find near-optimal solutions for the problem. We conduct extensive computational experiments to assess the performance of the algorithms.

APA, Harvard, Vancouver, ISO, and other styles

24

Lin, I.-sheng, and 林宜陞. "The Optimization of Fuzzy Membership Functions Using the Techniques of Genetic Algorithm and Reinforcement Learning for Vision-based Mobile Robots." Thesis, 2012. http://ndltd.ncl.edu.tw/handle/19082087224871620158.

Full text

Abstract:

碩士
國立高雄大學
電機工程學系碩士班
100
Fuzzy control is one of the popular control methods for intelligent robots. Usually, the membership functions associated with the fuzzy control rules are defined by domain experts heuristically. It usually takes a lot of time to adjust the fuzzy membership functions before they can be used practically. Machine learning techniques are adopted in this study for the optimization of fuzzy membership functions. Typical machine learning methods can be divided into two classes, the on-line learning and the off-line learning. On-line learning methods perform on physical machines and practical environments, whereas off-line learning methods run in computer-simulated environments. For robot applications, thousands of on-line learning iterations may be needed that take a long learning time and may damage the robot hardware. The results from off-line learning methods may not be applicable on a real robot due to the ideal assumptions, but unreality, of the simulators. The study proposed a two-stage learning approach. In the first stage, the genetic algorithm is used to perform off-line learning for deriving the best membership functions in a simulator. In the second stage, the fuzzy membership functions obtained from the simulator are on-line adjusted on a real robot using the techniques of reinforcement learning. The learning outcomes from the first stage are not practical but reasonable and they can be adjusted with fewer iterations of on-line learning in the second stage. In this manner, the overall learning time can be reduced with less damage on the robot hardware. The proposed method was verified on a real, wheel-type vision-based, robot. The experimental results show that the proposed method effectively produce optimal fuzzy membership functions for the control of a real robot.

APA, Harvard, Vancouver, ISO, and other styles

25

PAN, CHIEN-CHUN, and 潘建均. "A Remedial Learning System Based on a Genetic Algorithm and a Concept Map ── A Case Study of Object-Oriented Programming Language of Java." Thesis, 2018. http://ndltd.ncl.edu.tw/handle/3qjp27.

Full text

Abstract:

碩士
國立高雄師範大學
軟體工程與管理學系
106
Learning object-oriented programming (OOP) plays a very important role in computer science education. How to detect learners’ weakness of OOP concepts and therefore to lessen these weaknesses for learners remains an important issue. This study proposes an algorithm called Genetic Algorithm With Concept Map (GAWCM) to build a remedial learning system for a college level OOP course using a Genetic Algorithm (GA) and a concept map. Compared with linear optimization methods, GAs spend less computational time to find optimal or near-optimal solutions. Briefly, this study combines a GA and a concept map to select a set of remedial learning units from a repository according to learners' weaknesses of OOP concepts in a pre-test. The concept map serves to represent the knowledge structure of OOP, and the GA serves to select appropriate remedial learning units from the repository through its evolutionary computing. This study also conducted a simulation and the data were taken from a previous experiment from a national university in southern Taiwan. To validate the performance of the GAWCM, this study compared three different methods: the GAWCM, a GA, and a random method. The simulation results show that the results of GAWCM are better than those of the other two.

APA, Harvard, Vancouver, ISO, and other styles

26

Lockett, Alan Justin. "General-purpose optimization through information maximization." Thesis, 2012. http://hdl.handle.net/2152/ETD-UT-2012-05-5459.

Full text

Abstract:

The primary goal of artificial intelligence research is to develop a machine capable of learning to solve disparate real-world tasks autonomously, without relying on specialized problem-specific inputs. This dissertation suggests that such machines are realistic: If No Free Lunch theorems were to apply to all real-world problems, then the world would be utterly unpredictable. In response, the dissertation proposes the information-maximization principle, which claims that the optimal optimization methods make the best use of the information available to them. This principle results in a new algorithm, evolutionary annealing, which is shown to perform well especially in challenging problems with irregular structure.
text

APA, Harvard, Vancouver, ISO, and other styles

27

Taati, BABAK. "Generation and Optimization of Local Shape Descriptors for Point Matching in 3-D Surfaces." Thesis, 2009. http://hdl.handle.net/1974/5107.

Full text

Abstract:

We formulate Local Shape Descriptor selection for model-based object recognition in range data as an optimization problem and offer a platform that facilitates a solution. The goal of object recognition is to identify and localize objects of interest in an image. Recognition is often performed in three phases: point matching, where correspondences are established between points on the 3-D surfaces of the models and the range image; hypothesis generation, where rough alignments are found between the image and the visible models; and pose refinement, where the accuracy of the initial alignments is improved. The overall efficiency and reliability of a recognition system is highly influenced by the effectiveness of the point matching phase. Local Shape Descriptors are used for establishing point correspondences by way of encapsulating local shape, such that similarity between two descriptors indicates geometric similarity between their respective neighbourhoods. We present a generalized platform for constructing local shape descriptors that subsumes a large class of existing methods and allows for tuning descriptors to the geometry of specific models and to sensor characteristics. Our descriptors, termed as Variable-Dimensional Local Shape Descriptors, are constructed as multivariate observations of several local properties and are represented as histograms. The optimal set of properties, which maximizes the performance of a recognition system, depend on the geometry of the objects of interest and the noise characteristics of range image acquisition devices and is selected through pre-processing the models and sample training images. Experimental analysis confirms the superiority of optimized descriptors over generic ones in recognition tasks in LIDAR and dense stereo range images.
Thesis (Ph.D, Electrical & Computer Engineering) -- Queen's University, 2009-09-01 11:07:32.084

APA, Harvard, Vancouver, ISO, and other styles

Dissertations / Theses on the topic 'Genetic algorithm based learning algorithm (GABL)'

Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles