Mutual learning is a machine learning algorithm where multiple machine learning algorithms share knowledge among themselves to improve themselves.The utilization of mutual learning algorithms can effectively enhance the efficiency of machine learning and neural networks within a multiagent system.This approach is particularly useful in scenarios where the system cannot be adequately trained with a large dataset.By exchanging data in a dynamic teacher-student network system, mutual learning can result in efficient learning outcomes.Typically, a large network serves as a static teacher and transfers data to smaller networks, referred to as student networks, to improve their efficiency.In this study, we aim to demonstrate that two small networks can dynamically alternate between the roles of teacher and student to share knowledge, resulting in improved efficiency for both networks.To exemplify this concept, we apply a mutual learning algorithm using convolutional neural networks (CNNs) and Support Vector Machine (SVM) to accurately identify the kidney diseasescyst, tumor and stone using image classification algorithm.

Mutual learning algorithm can be an efficient mechanism for improving the machine learning and neural network efficiency in a multi-agent system. Specifically, in many cases, where the system cannot be trained using a big training dataset, the data exchange in teacher-student network system can lead to efficient learning. Usually, in mutual learning algorithms, a big network plays the role of a static teacher and passes the data to smaller networks, known as student networks, to improve the efficiency of the latter. In this paper, we will show that two small networks can dynamically play the changing roles of teacher and student to share their knowledge and hence, the efficiency of both the networks improve simultaneously. We demonstrate the concept and the proposed mutual learning algorithm using convolutional neural networks (CNNs) to recognize the benchmark Modified National Institute of Standards and Technology (MNIST) hand-writing dataset.

In two earlier papers presented at the 2019 and 2020 American Control Conferences, the concept of “Mutual Learning” was introduced by the authors and applied to learning in static and dynamic stochastic environments. In this paper, we extend the concept of mutual learning to optimization. Two agents attempting to optimize the same performance index “learn” from each other to reach the solution more efficiently. Since optimization is a well investigated mathematical area in systems theory, it is particularly well suited to the original objective of the authors to study “Mutual Learning” in a systems theoretic framework.The two agents involved in mutual learning can use any of the methods well-known in the literature to optimize the given function. The initial conditions and the period over which the optimization is carried out, may be different for the two agents before they communicate with each other for the first time. The principal conclusion of the paper is that mutual learning should be viewed as a general research area, and not as a specific procedure used in different system theoretic problems.

Mutual learning is an emerging technique for allowing intelligent systems to learn from each other, giving rise to improved performance. In this paper, we explore mutual reinforcement learning between systems which use very different learning algorithms. In particular, we present an algorithm which allows two agents, one using Q-learning and another using adaptive dynamic programming, to share learned knowledge. We discuss how these agents negotiate the relative importance of knowledge they receive from other agents, and we present results that show how this affects the learning process.

Machine learning algorithms and traditional data mining process usually require a large volume of data to train the algorithm-specific models, with little or no user feedback during the model building process. Such a "big data" based automatic learning strategy is sometimes unrealistic for applications where data collection or processing is very expensive or difficult, such as in clinical trials. Furthermore, expert knowledge can be very valuable in the model building process in some fields such as biomedical sciences. In this paper, we propose a new visual analytics approach to interactive machine learning and visual data mining. In this approach, multi-dimensional data visualization techniques are employed to facilitate user interactions with the machine learning and mining process. This allows dynamic user feedback in different forms, such as data selection, data labeling, and data correction, to enhance the efficiency of model building. In particular, this approach can significantly reduce the amount of data required for training an accurate model, and therefore can be highly impactful for applications where large amount of data is hard to obtain. The proposed approach is tested on two application problems: the handwriting recognition (classification) problem and the human cognitive score prediction (regression) problem. Both experiments show that visualization supported interactive machine learning and data mining can achieve the same accuracy as an automatic process can with much smaller training data sets.

Dynamic zero-sum games are a model of multiagent decision-making that has been well-studied in the mathematical game theory literature. In this paper, we derive a sufficient condition for the existence of a solution to this problem, and then proceed to discuss various reinforcement learning strategies to solve such a dynamic game in the presence of uncertainty where the game matrices at various states as well as the transition probabilities between the states under different agent actions are unknown. A novel algorithm, based on heterogeneous games of learning automata (HEGLA), as well as algorithms based on model-based and model-free reinforcement learning, are presented as possible approaches to learning the solution Markov equilibrium policies when they are assumed to satisfy the sufficient conditions for existence. The HEGLA algorithm involves automata simultaneously playing zero-sum games with some automata and identical pay-off games with some other automata. Simulation studies are reported to complement the theoretical and algorithmic discussions.

Multidisciplinary Optimization (MDO) is one of the most popular techniques in aerospace engineering, where the system is complex and includes the knowledge from multiple fields. However, according to the best of our knowledge, MDO has not been widely applied in decentralized reinforcement learning (RL) due to the `unknown' nature of the RL problems. In this work, we apply the MDO in decentralized RL. In our MDO design, each learning agent uses system identification to closely approximate the environment and tackle the `unknown' nature of the RL. Then, the agents apply the MDO principles to compute the control solution using Monte Carlo and Markov Decision Process techniques. We examined two options of MDO designs: the multidisciplinary feasible and the individual discipline feasible options, which are suitable for multi-agent learning. Our results show that the MDO individual discipline feasible option could successfully learn how to control the system. The MDO approach shows better performance than the completely decentralization and centralization approaches.

In this paper, we explore the capability of selectively decentralized Q-learning approach in learning how to optimally stabilize control systems, as compared to the centralized approach. We focus on problems in which the systems are completely unknown except the possible domain knowledge that allow us to decentralize into subsystems. In selective decentralization, we explore all of the possible communication policies among subsystems and use the cumulative gained Q-value as the metric to decide which decentralization scheme should be used for controlling. The results show that the selectively decentralized approach not only stabilizes the system faster but also shows superior converging speed on gained Q-value in different systems with different interconnection strength. In addition, the selectively decentralized converging time does not seem to grow exponentially with the system dimensionality. Practically, this fact implies that the selectively decentralized Q-learning could be used as an alternative approach in large-scale unknown control system, where in theory, the Hamilton-Jacobi-Bellman-equation approach is difficult to derive the close-form solution.

Determining the optimal design of a watershed is a highly subjective process which involves the consideration of many distinct factors by several different stakeholder groups. We describe additional functionality for our watershed planning system, called WRESTORE (Watershed REstoration Using Spatio-Temporal Optimization of REsources) (http://wrestore.iupui.edu), where stakeholders can collaboratively optimize best management practices on to the watershed. WRESTORE utilizes the USDA's public domain Soil and Water Assessment Tool hydrologic model for watershed simulations. Reinforcement learning and interactive genetic algorithms are applied for the search process. The new functionality described is a user modeling component that develops a computational model of a user's decision-making, based on real-time user-provided ratings for a subset of possible designs. The user modeling task utilizes neural network approaches, such as deep learning. We believe the originality of our approach centers on integrating user models in to the hydrological decision support process. This paper thus has three objectives: (i) outline current work in user modeling and watershed design, (ii) describe our system for interactive optimization of watershed design, and (iii) describe our work on implementing accurate and stable user predictive models to boost optimization performance.

In this paper, we explore the capability of selective decentralization in improving the control performance for unknown large-scale systems using model-based approaches. In selective decentralization, we explore all of the possible communication policies among subsystems and show that with the appropriate switching among the resulting multiple identification models (with corresponding communication policies), such selective decentralization significantly outperforms a centralized identification model when the system is weakly interconnected, and performs at least equivalent to the centralized model when the system is strongly interconnected. To derive the sub-optimal control, our control design include two phases. First, we apply system identification to train the approximation model for the unknown system. Second, we find the suboptimal solution of the Halminton-Jacobi-Bellman (HJB) equation to derive the suboptimal control. In linear systems, the HJB equation transforms to the well-solved Riccati equation with closed-form solution. In nonlinear systems, we discretize the approximation model in order to acquire the control unit by using dynamic programming methods for the resulting Markov Decision Process (MDP). We compare the performance among the selective decentralization, the complete decentralization and the centralization in our two-phase control design. Our results show that selective decentralization outperforms the complete decentralization and the centralization approaches when the systems are completely decoupled or strongly interconnected.

We have developed a web-based, interactive, watershed planning system called WRESTORE (Watershed Restoration Using Spatio-Temporal Optimization of Resources) (http://wrestore.iupui.edu) that allows stake-holder communities to participate in a democratic, collaborative form of optimization process for designing best management practices (BMPs) on their landscape, while also optimizing based on subjective, qualitative landowners' criteria beyond the usual socio-economic, physical, and ecological criteria. This system utilizes multiple advanced computational approaches including the SWAT (Soil and Water Assessment Tool) hydrologic model for watershed simulations, interactive genetic algorithms and reinforcement-based machine learning algorithms for search and optimization, and deep learning artificial neural networks for user modeling, within an encompassing human-computer interaction framework. A substantial user study of the WRESTORE system was conducted recently involving multiple real stakeholders varying from consultants, government officials, watershed alliance members, etc., with the objective of gaining insight about WRESTORE'S usability and utility. In particular focus was the user modeling component that develops a computational model of a user's preferences and criteria, based on real-time user-provided ratings for a subset of possible designs (similar to the idea of user profiling commonly done for human-computer interaction systems). The user model constructed based on the real user's personalized feedbacks can then be used to influence the automated search and optimization for BMP alternatives in WRESTORE. In this paper, we describe the methods developed for user modeling for interactive optimization, and the experimental set-up as well as results with real user studies. These results clearly demonstrate that development of user models for such personalized, interactive optimization is both feasible and valuable for developing community-based computational water sustainability solutions.

User modelling is one of the prominent research fields in information retrieval systems. In this paper, we model user's preferences and search criteria using an NN (Neural Network) to solve a multiobjective optimization problem specific to environmental planning systems. We argue that some NP hard problems cannot be solved alone either by a human or by a computer. Human participation in automated search is one way of combining human intuition with algorithmic search to solve such problems. However, even humans have some limitations for participation in that they cannot participate in search completely because of human fatigue. To overcome this, in our approach, an NN tries to model the user's rating criteria and preferences to help the user in rating large set of designs. Although training an NN with limited data is not always feasible, there are many situations where a simple modelling technique (e.g., linear/quadratic mapping) works better if the learning data set is small. In this paper we attempt to get more accuracy of the NN by generating data using other linear/non-linear techniques that fills the gap created by lack of sufficient training data. Also, we provided the architectural design of an HPC based framework we have proposed and compared the performance of the NN with fuzzy logic and other linear/non-linear user modelling techniques for the environmental resources optimization problem.

Mining frequent patterns from a hidden dataset is an important task with 43 various real-life applications. In this research, we propose a solution to this problem that is based on Markov Chain Monte Carlo (MCMC) sampling of frequent patterns. Instead of returning all the frequent patterns, the proposed paradigm returns a small set of randomly selected patterns so that the clandestinity of the dataset can be maintained. Our solution also allows interactive sampling, so that the sampled patterns can fulfill the user's requirement effectively. We show experimental results from several real life datasets to validate the capability and usefulness of our solution; in particular, we show examples that by using our proposed solution, an eCommerce marketplace can allow pattern mining on user session data without disclosing the data to the public; such a mining paradigm helps the sellers of the marketplace, which eventually boost the marketplace's own revenue.

In this paper, we use identical-payoff games of reinforcement learning agents as a framework to solve complex multi-criteria optimization problem of watershed management. Multiple analytical criteria are used to assess design decisions for creating a distributed network of wetlands in the watershed. Decentralized game algorithms of reinforcement learning agents as well as a genetic algorithm based method are used for the analysis. Simulation studies are presented which compare the efficiency of the reinforcement learning approaches with a multi-objective genetic algorithm-based approach.

SUMMARY Osteoporosis is characterized by reduced bone mass and debilitating fractures and is likely to reach epidemic proportions. Because of the vigorous research taking place in fields related to osteoporosis, bone biologists are overwhelmed by the amount of literature being generated on a regular basis. This problem can be alleviated by inferring and extracting novel relationships among biological entities appearing in the biological literature. With the development of large online publicly available databases of biological literature, such an approach becomes even more appealing. The novel relationships between biological terms thus discovered constitute new hypotheses that can be verified using experiments. This paper presents a novel method called multilevel text mining for the extraction of potentially meaningful biological relationships. Multilevel mining uses transitive maximum flow graph analysis coupled with set combination operations of union and intersection. Set operators are applied along and across the paths of a transitive flow graph to combine the data. In the first level of the multilevel mining process, protein domain names are used. Novel relationships between domains are extracted by the transitive text mining analysis. In the second level, these newly discovered relationships are used to extract relevant protein names. Set operators are used in various combinations to obtain different sets of results. Copyright © 2011 John Wiley & Sons, Ltd.

We discuss the application of indirect learning methods in zero-sum and identical payoff learning automata games. We propose a novel decentralized version of the well-known pursuit learning algorithm. Such a decentralized algorithm has significant computational advantages over its centralized counterpart. The theoretical study of such a decentralized algorithm requires the analysis to be carried out in a nonstationary environment. We use a novel bootstrapping argument to prove the convergence of the algorithm. To our knowledge, this is the first time that such analysis has been carried out for zero-sum and identical payoff games. Extensive simulation studies are reported, which demonstrate the proposed algorithm's fast and accurate convergence in a variety of game scenarios. We also introduce the framework of partial communication in the context of identical payoff games of learning automata. In such games, the automata may not communicate with each other or may communicate selectively. This comprehensive framework has the capability to model both centralized and decentralized games discussed in this paper.

Osteoporosis, which is characterized by reduced bone mass and debilitating fractures, may reach epidemic proportions with the aging of the US population. The intensity of research in this field of study is reflected by the facts that The American Society of Bone and Mineral Research has a membership of nearly 4,000 physicians, clinical investigators, and basic research scientists from over fifty countries and that NIH is expected to spend over 200 million dollars on osteoporosis research alone in 2010. Bone biologists may be overwhelmed by the amount of literature constantly being generated, thus the identification and extraction of existing and novel relationships among biological entities or terms appearing in the biological literature is an ongoing problem. The problem has become more pressing with the development of large online publicly available databases of biological literature. Extraction and visualization of relationships between biological entities appearing in these databases offers the opportunity of keeping researchers up-to-date in their research domain. This may be achieved through helping them visualize possible biological pathways and by generating likely new hypotheses concerning novel interactions through methods such as transitive closure network flow. All generated predictions can be verified against already existing data, and possible new relationships can be verified against experiment. This paper presents a method for the extraction and visualization of potentially meaningful relationships.

In this paper, we describe a framework for solving computationally hard, distributed function optimization problems using reinforcement learning techniques. In particular, we model a function optimization problem as an identical payoff game played by a team of reinforcement learning agents. The team performs a stochastic search through the domain space of the parameters of the function. However, current game learning algorithms suffer from significant memory requirement, significant communication overhead and slow convergence. To alleviate these problems, we present novel decentralized and partially decentralized reinforcement learning algorithms for the team. Simulation results are presented for the NP-Hard sensor subset selection problem to show that the agents learn locally optimal parameter values and illustrate the advantages of the proposed algorithms.

Breast cancer and osteoporosis are two most common diseases in postmenopausal women. Both diseases are multi-factorial and involve complex interactions of many genes. Since it is very difficult to review all published papers manually to understand interaction between genes pertaining to these two diseases, we employed text mining system which is an automated approach to search for these gene interactions. Two gene lists were first constructed. The first one contained genes that may be involved in breast cancer, and the second one included those that may be involved in osteoporosis. Potential transitive or indirect associations between two gene terms were determined using transitive closure on the direct associations extracted on the basis of co-occurrence of gene terms in the abstracts. The transitive associations were ranked using a graph-based weight scoring algorithm. With this scoring method, the top 10 gene pairs that are most likely associated with these two diseases were found to be p53/osteocalcin, VEGF/IGF-1, BRAC1/osteocalcin, p53/IL-6, IGFBP3/ESR-alpha, COMT/CYP1A1, p53/OPG, VEGF/OPG, and IGFBP3/RANK. This study also revealed a potential link of P53 in both diseases. Further investigations are required to characterize and confirm this association.

In this paper, we introduce reinforcement learning as a methodology to solve complex multi-criteria optimization problems for ground water monitoring. Multiple analytical criteria are used to assess design decisions and human feedback is simulated by adding random noise. Different learning automata based reinforcement learning methods as well as a genetic algorithm based method are used in experimental studies, which demonstrate the efficiency of reinforcement learning approaches.

There has been a considerable amount of recent research in extraction of various kinds of binary associations (e.g., gene-gene, gene-protein, protein-protein, etc) using different text mining approaches. However, an important aspect of such associations is identifying the context in which such associations occur (e.g., "gene A activates protein B in the context of disease C in organ D under the influence of chemical E"). Such contexts can be represented appropriately by a multi-way relationship involving more than two objects rather than usual binary relationships. Such multi-way relations naturally lead to a hyper-graph representation of the knowledge. The hyper-graph based knowledge extraction from biological literature represents a computationally difficult problem due to its combinatorial nature. In this paper, we compare two different approaches to such hyper-graph extraction: one based on an exhaustive enumeration of all hyper-edges and the other based on an extension of the well-known A Priori algorithm.

Abstract Information filtering is a technique to identify, in large collections, information that is relevant according to some criteria (e.g., a user's personal interests, or a research project objective). As such, it is a key technology for providing efficient user services in any large‐scale information infrastructure, e.g., digital libraries. To provide large‐scale information filtering services, both computational and knowledge management issues need to be addressed. A centralized (single‐agent) approach to information filtering suffers from serious drawbacks in terms of speed, accuracy, and economic considerations, and becomes unrealistic even for medium‐scale applications. In this article, we discuss two distributed (multi‐agent) information filtering approaches, that are distributed with respect to knowledge or functionality, to overcome the limitations of single‐agent centralized information filtering. Large‐scale experimental studies involving the well‐known TREC data set are also presented to illustrate the advantages of distributed filtering as well as to compare the different distributed approaches.

Abstract There has been considerable interest in recent years in providing automated information services, such as information classification, by means of a society of collaborative agents. These agents augment each other's knowledge structures (e.g., the vocabularies) and assist each other in providing efficient information services to a human user. However, when the number of agents present in the society increases, exhaustive communication and collaboration among agents result in a large communication overhead and increased delays in response time. This paper introduces a method to achieve selective interaction with a relatively small number of potentially useful agents, based on simple agent modeling and acquaintance lists. The key idea presented here is that the acquaintance list of an agent, representing a small number of other agents to be collaborated with, is dynamically adjusted. The best acquaintances are automatically discovered using a learning algorithm, based on the past history of collaboration. Experimental results are presented to demonstrate that such dynamically learned acquaintance lists can lead to high quality of classification, while significantly reducing the delay in response time.

Today a massive amount of information available on the WWW often makes searching for information of interest a long and tedious task. Chasing hyperlinks to find relevant information may be daunting. To overcome such a problem, a learning system, cognizant of a user's interests, can be employed to automatically search for and retrieve relevant information by following appropriate hyperlinks. In this paper, we describe the design of such a learning system for automated Web navigation using adaptive dynamic programming methods. To improve the performance of the learning system, we introduce the notion of multiple model-based learning agents operating in parallel, and describe methods for combining their models. Experimental results on the WWW navigation problem are presented to indicate that combining multiple learning agents, relying on user feedback, is a promising direction to improve learning speed in automated WWW navigation.

MOTIVATION: As biomedical researchers are amassing a plethora of information in a variety of forms resulting from the advancements in biomedical research, there is a critical need for innovative information management and knowledge discovery tools to sift through these vast volumes of heterogeneous data and analysis tools. In this paper we present a general model for an information management system that is adaptable and scalable, followed by a detailed design and implementation of one component of the model. The prototype, called BioSifter, was applied to problems in the bioinformatics area. RESULTS: BioSifter was tested using 500 documents obtained from PubMed database on two biological problems related to genetic polymorphism and extracorporal shockwave lithotripsy. The results indicate that BioSifter is a powerful tool for biological researchers to automatically retrieve relevant text documents from biological literature based on their interest profile. The results also indicate that the first stage of information management process, i.e. data to information transformation, significantly reduces the size of the information space. The filtered data obtained through BioSifter is relevant as well as much smaller in dimension compared to all the retrieved data. This would in turn significantly reduce the complexity associated with the next level transformation, i.e. information to knowledge.

In recent years, systems consisting of multiple modular neural networks have attracted substantial interest in the neural networks community because of various advantages they offer over a single large monolithic network. In this paper, we propose two basic feature decomposition models (namely, parallel model and tandem model) in which each of the neural network modules processes a disjoint subset of the input features. A novel feature decomposition algorithm is introduced to partition the input space into disjoint subsets solely based on the available training data. Under certain assumptions, the approximation error due to decomposition can be proved to be bounded by any desired small value over a compact set. Finally, the performance of feature decomposition networks is compared with that of a monolithic network in real world bench mark pattern recognition and modeling problems.

Information services such as searching, retrieval, and filtering are playing a dominant role in our life during the current information age. One critical functionality of these information services is to obtain effective classification for input documents. Thesaurus(vocabulary)-based document representation followed by clustering constitutes a popular approach to document classification. However, two alternatives exist to construct the information classificationsystem. The first one uses a single, monolithic, huge thesaurus and classifies all documents by one centralized machine. The second one exploits distributed computing environmentsby allowing multiple agents with small thesauri to collaborate with each other over a computer network. The objective of this paper is to compare these two approaches (i.e., single-agent and multi-agent) in terms of various criteria including response time, quality of classification, and economic/privacy considerations. Two experimental studies, involving classification of Computer Science and Medline documents, are presented to compare the performanceof a single-agent system with that of a multi-agent system in real world settings. These results indicate that a collaborative multi-agent system constitutes a attractive methodologyfor classifying a large volume of information efficiently, when the thesaurus is large.

The NARMA model is an exact representation of the input-output behavior of finite-dimensional nonlinear discrete-time dynamical systems in a neighborhood of the equilibrium state. However, it is not convenient for purposes of adaptive control using neural networks due to its nonlinear dependence on the control input. Hence, quite often, approximate methods are used for realizing the neural controllers to overcome computational complexity. In this paper, we introduce two classes of models which are approximations to the NARMA model, and which are linear in the control input. The latter fact substantially simplifies both the theoretical analysis as well as the practical implementation of the controller. Extensive simulation studies have shown that the neural controllers designed using the proposed approximate models perform very well, and in many cases even better than an approximate controller designed using the exact NARMA model. In view of their mathematical tractability as well as their success in simulation studies, a case is made in this paper that such approximate input-output models warrant a detailed study in their own right.

In information-filtering environments, uncertainties associated with changing interests of the user and the dynamic document stream must be handled efficiently. In this article, a filtering model is proposed that decomposes the overall task into subsystem functionalities and highlights the need for multiple adaptation techniques to cope with uncertainties. A filtering system, SIFTER, has been implemented based on the model, using established techniques in information retrieval and artificial intelligence. These techniques include document representation by a vector-space model, document classification by unsupervised learning, and user modeling by reinforcement learning. The system can filter information based on content and a user's specific interests. The user's interests are automatically learned with only limited user intervention in the form of optional relevance feedback for documents. We also describe experimental studies conducted with SIFTER to filter computer and information science documents collected from the Internet and commercial database services. The experimental results demonstrate that the system performs very well in filtering documents in a realistic problem setting.

Article Free Access Share on Detection of shifts in user interests for personalized information filtering Authors: W. Lam Department of Management Sciences, S306 Pappajohn Building, The University of Iowa, Iowa City, Iowa Department of Management Sciences, S306 Pappajohn Building, The University of Iowa, Iowa City, IowaView Profile , S. Mukhopadhyay Computer and Information Science, Purdue University School of Science at Indianapolis, 723 W. Michigan St. SL280, Indianapolis, IN Computer and Information Science, Purdue University School of Science at Indianapolis, 723 W. Michigan St. SL280, Indianapolis, INView Profile , J. Mostafa School of Library and Information Science, Indiana University, Bloomington, IN School of Library and Information Science, Indiana University, Bloomington, INView Profile , M. Palakal Computer and Information Science, Purdue University School of Science at Indianapolis, 723 W. Michigan St. SL280, Indianapolis, IN Computer and Information Science, Purdue University School of Science at Indianapolis, 723 W. Michigan St. SL280, Indianapolis, INView Profile Authors Info & Claims SIGIR '96: Proceedings of the 19th annual international ACM SIGIR conference on Research and development in information retrievalAugust 1996 Pages 317–325https://doi.org/10.1145/243199.243279Published:18 August 1996Publication History 36citation915DownloadsMetricsTotal Citations36Total Downloads915Last 12 Months40Last 6 weeks5 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

Neural networks with different architectures have been successfully used for the identification and control of a wide class of nonlinear systems. The problem of rejection of input disturbances, when such networks are used in practical problems is considered. A large class of disturbances, which can be modeled as the outputs of unforced linear or nonlinear dynamic systems, is treated. The objective is to determine the identification model and the control law to minimize the effect of the disturbance at the output. In all cases, the method used involves expansion of the state space of the disturbance-free plant in an attempt to eliminate the effect of the disturbance. Several stages of increasing complexity of the problem are discussed in detail. Theoretical justification is provided for the existence of solutions to the problem of complete rejection of the disturbance in special cases. This provides the rationale for using similar techniques in situations where such theoretical analysis is not available.

Associative learning is investigated using neural networks and concepts based on learning automata. The behavior of a single decision-maker containing a neural network is studied in a random environment using reinforcement learning. The objective is to determine the optimal action corresponding to a particular state. Since decisions have to be made throughout the context space based on a countable number of experiments, generalization is inevitable. Many different approaches can be followed to generate the desired discriminant function. Three different methods which use neural networks are discussed and compared. In the most general method, the output of the network determines the probability with which one of the actions is to be chosen. The weights of the network are updated on the basis of the actions and the response of the environment. The extension of similar concepts to decentralized decision-making in a context space is also introduced. Simulation results are included. Modifications in the implementations of the most general method to make it practically viable are also presented. All the methods suggested are feasible and the choice of a specific method depends on the accuracy desired as well as on the available computational power.

The terrain acquisition problem is formulated as that of continuous motion planning, and no constraints are imposed on obstacle geometry. Two algorithms are described for acquiring planar terrains with obstacles of arbitrary shape. Estimates of the algorithm performance are derived as upper bounds on the lengths of generated paths.>

A cooperative-game-playing learning automata model is presented for a complex nonlinear associative task, namely, learning of Boolean functions. The unknown Boolean function is expressed in terms of minterms, and a team of automata is used to learn the minterms present in the expansion. Only noisy outputs of the Boolean function are assumed to be available for the team of automata that use a variation of the rapidly converging estimator learning algorithm called the pursuit algorithm. A divide-and-conquer approach is proposed to overcome the storage and computational problems of the pursuit algorithm. Extensive simulation experiments have been carried out for six-input Boolean tasks. The main advantages offered by the model are generality, proof of convergence, and fast learning.>