By Topic

Evolving and Adaptive Intelligent Systems (EAIS), 2012 IEEE Conference on

Date 17-18 May 2012

Filter Results

Displaying Results 1 - 25 of 45
  • 2012 IEEE Conference on Evolving and Adaptive Intelligent Systems [Front cover]

    Page(s): c1
    Save to Project icon | Request Permissions | PDF file iconPDF (401 KB)  
    Freely Available from IEEE
  • 2012 IEEE Conference on Evolving and Adaptive Intelligent Systems [title page]

    Page(s): i
    Save to Project icon | Request Permissions | PDF file iconPDF (222 KB)  
    Freely Available from IEEE
  • Proceedings of 2012 IEEE Conference on Evolving and Adaptive Intelligent Systems - (IEEE EAIS12) [Copyright notice]

    Page(s): ii
    Save to Project icon | Request Permissions | PDF file iconPDF (40 KB)  
    Freely Available from IEEE
  • Preface

    Page(s): iii - iv
    Save to Project icon | Request Permissions | PDF file iconPDF (55 KB)  
    Freely Available from IEEE
  • Program committee

    Page(s): v
    Save to Project icon | Request Permissions | PDF file iconPDF (27 KB)  
    Freely Available from IEEE
  • Organizing committee

    Page(s): vi
    Save to Project icon | Request Permissions | PDF file iconPDF (28 KB)  
    Freely Available from IEEE
  • 2012 IEEE Conference on Evolving and Adaptive Intelligent Systems - Program

    Page(s): vii - ix
    Save to Project icon | Request Permissions | PDF file iconPDF (1290 KB)  
    Freely Available from IEEE
  • Table of contents

    Page(s): x - xii
    Save to Project icon | Request Permissions | PDF file iconPDF (401 KB)  
    Freely Available from IEEE
  • On-line active learning based on enhanced reliability concepts

    Page(s): 1 - 6
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (897 KB) |  | HTML iconHTML  

    In this paper, we present a new methodology for conducting active learning in a single-pass on-line learning context, thus reducing the annotation effort for operators by selecting the most informative samples, i.e. those ones helping incremental, evolving classifiers most to improve their own predictive performance. Our approach will be based on certainty-based sample selection in connection with version-space reduction approach. Therefore, two new concepts regarding classifier's reliability in its predictions will be investigated and developed in connection with evolving fuzzy classifiers: conflict and ignorance. Conflict models the extent to which a new query point lies in the conflicting region between two or more classes. Ignorance represents the extent to which the new query point appears in an unexplored region of the feature space. The results based on real-world streaming classification data will show a stable high predictive quality of our approach, despite the fact that the requested number of class labels is decreased by up to 90%. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A property of learning chunk data using incremental kernel principal component analysis

    Page(s): 7 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (568 KB) |  | HTML iconHTML  

    An incremental learning algorithm of Kernel Principal Component Analysis (KPCA) called Chunk Incremental KPCA (CIKPCA) has been proposed for online feature extraction in pattern recognition. CIKPCA can reduce the number of times to solve the eigenvalue problem compared with the conventional incremental KPCA when a small number of data are simultaneously given as a stream of data chunks. However, our previous work suggests that the computational costs of the independent data selection in CIKPCA could dominate over those of the eigenvalue decomposition when a large chunk of data are given. To verify this, we investigate the influence of the chunk size to the learning time in CIKPCA. As a result, CIKPCA requires more learning time than IKPCA unless a large chunk of data are divided into small chunks (e.g., less than 50). View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Self-adaptive mechanism for distributed computing

    Page(s): 11 - 16
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1274 KB) |  | HTML iconHTML  

    This paper proposes a framework for adapting software components for a disaggregated system, which dynamically composes devices, e.g., displays, keyboard, and mice, which do not attached to the same computer into a virtual computer over a distributed system, including ubiquitous/pervasive computing systems. It introduces the notions of differentiation and dedifferentiation in cellular slime molds into real distributed systems, including disaggregated systems. When software components delegates a function to another component coordinating with it, if the former has the function, this function becomes less-developed and the latter's function becomes well-developed. The framework was constructed as a middleware system and allowed us to define agents as Java objects written in JavaBean. We present several evaluations of the framework in a distributed system instead of any simulation-based systems and describes a practical application. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Online learning with kernels in classification and regression

    Page(s): 17 - 22
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (645 KB) |  | HTML iconHTML  

    New optimization models and algorithms for online learning with kernels (OLK) in classification and regression are proposed in a Reproducing Kernel Hilbert Space (RKHS) by solving a constrained optimization model. The “forgetting” factor in the model makes it possible that the memory requirement of the algorithm can be bounded as the learning process continues. The applications of the proposed OLK algorithms in classification and regression show their effectiveness in comparing with the state of art algorithms. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An enhanced approach for evolving participatory learning fuzzy modeling

    Page(s): 23 - 28
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (424 KB) |  | HTML iconHTML  

    Evolving participatory learning (ePL) modeling joins the concepts of participatory learning and evolving fuzzy systems. It uses data streams to continuously adapt the structure and functionality of fuzzy models. This paper proposes an enhanced version of the ePL approach, called ePL+, which includes both an utility measure as a mechanism to shrink rule bases, and a variable zone of influence of clusters. These features are useful in fuzzy rule-based modeling to construct the fuzzy rules. Computational experiments with the classic Mackey-Glass and Box & Jenkins benchmarks are conducted to compare the performance of the ePL+ approach with state of the art alternative fuzzy modeling methods and double exponential smoothing technique. The results show the high capability of ePL+ to model time series; it produces accurate results with a robust, flexible and autonomous algorithm. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive evolution strategy for robotic manipulation

    Page(s): 29 - 34
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (867 KB) |  | HTML iconHTML  

    Teaching a mobile robot to complete a task and to reproduce it is possible, but as the robot tries to replicate actions natural events as a wheel-slide would feed in inaccuracies on the localization of the robot mobile base, and it may be difficult to succeed replicating. Robot tasks can be represented as trajectories compound by a series of poses and movements. We propose an algorithm for adapting manipulation trajectories to different initial conditions from those of the learned assignment. The adaptation is achieve by optimizing in position, orientation and energy conservation. Manipulation paths generated can achieve optimal performance sometimes even improving original path smoothness. The approach is builded over the basis of Evolution Strategies(ES), and only uses forward kinematics permitting to avoid all the inconveniences that Inverse kinematics imply as well as convergence problems in singular kinematic configurations. Experimental results are presented to verify the algorithm. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Elastic adaptive ontology matching on evolving folksonomy driven environment

    Page(s): 35 - 40
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (679 KB) |  | HTML iconHTML  

    Semantic networks can simulate the human complex frames in reasoning process providing efficient association and inference mechanisms. Ontology can be used to fill the gap between human and computational intelligence for a task domain. For an evolving environment it is important to understand what knowledge is required for a task domain with an adaptive ontology matching. To reflect the evolving knowledge this paper considers ontologies based on folksonomies according to a new concept structure called “Folksodriven” to represent folksonomies. Folksonomies are a set of terms that a group of users tagged content without a controlled vocabulary. A Folksodriven Structure Network (FSN), built from the relations among the Folksodriven tags, is presented as a folksonomy tags suggestions for the user to solve the problems inherent in an uncontrolled vocabulary of the folksonomy. It was observed that the properties of the FSN depend mainly on the nature, distribution, size and the quality of the reinforcing Folksodriven tags (FD tags). So, the studies on the transformational regulation of the FD tags are regarded to be important for an adaptive folksonomies classifications in an evolving environment used by Intelligent Systems. This paper discuss the deformation exhibiting linear behavior on FSN based on folksonomy tags chosen by different user on web site resources, this is a topic which has not been well studied so far. The discussion shows that the linear elastic constitutive equation possesses some leaning for the investigation. A constitutive law on FSN is investigated towards a systematic mathematical analysis on stress analysis and equations of motion for an evolving ontology matching on an environment defined by the users' folksonomy choice. The adaptive ontology matching and the elastodynamics are merged to obtain what we can call the elasto-adaptative-dynamics methodology of the FSN. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Cognitively inspired classification for adapting to data distribution changes

    Page(s): 41 - 46
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1065 KB) |  | HTML iconHTML  

    In pattern classification, the test data is expected to lie in the domain covered by the training data. But in practical scenarios, this may not necessarily be true. To improve the adaptability, the classifier should be able to generalize well even when there are changes in the input distribution. This paper proposes a cognitively inspired classification framework based on rules and exemplars. It can generalize well even for samples falling outside the region covered by the training data. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A new fuzzy adaptive law with leakage

    Page(s): 47 - 50
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (733 KB) |  | HTML iconHTML  

    In this paper a novel adaptive law with leakage is presented. It is shown in the paper that the proposed adaptive law is a natural way to cope with the parasitic dynamics. Moreover, the value of the leakage parameter σ́ is directly related to the norm of the parasitic dynamics. This is the unique property of the proposed adaptive law since this is not the case when using σ-modification, switching σ-modification, or e1-modification. The boundedness of estimated parameters, the tracking error and all the signals in the system is proven if the leakage parameter σ́ satisfies certain condition. This means that the proposed adaptive law ensures the global stability of the system. A simulation example is given that illustrates the proposed approach. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An adaptive soft-sensor for non-destructive cement-based material testing, through the use of RBF networks

    Page(s): 51 - 56
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1202 KB) |  | HTML iconHTML  

    This paper presents the development of a soft-sensor receiving as inputs Pressure Stimulated Current (PSC) characteristics in order to predict a critical mechanical property of cement-based materials, in a non-destructive manner. The soft-sensor is based on a Radial Basis Function (RBF) network that starts with an empty hidden layer and evolves its structure and synaptic weights as new data become available. Results have shown that the proposed approach can be used successfully to evolve a predictive tool based on input-output data, whereas it is superior compared to other adaptive modeling techniques. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A procedural Long Term Memory for cognitive robotics

    Page(s): 57 - 62
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (956 KB) |  | HTML iconHTML  

    This paper provides some insights into the advantages of using a Long-Term Memory (LTM) for optimizing the adaptive learning capabilities of a cognitive robot in dynamic environments. Specifically, a procedural LTM that stores basic models and behaviours is included in the evolutionary-based Multilevel Darwinist Brain (MDB) cognitive architecture. The memory system is based on learning error stability and instability to detect if a model is candidate to enter the LTM or to be recovered. A LTM replacement strategy has been developed that is based on context detection using functional comparison of the models' response. The LTM elements are tested in theoretical functions and in a simulated example using the AIBO robot in a dynamic context with successful adaptive learning results. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive strategy for online gait learning evaluated on the polymorphic robotic LocoKit

    Page(s): 63 - 68
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1065 KB) |  | HTML iconHTML  

    This paper presents experiments with a morphology-independent, life-long strategy for online learning of locomotion gaits, performed on a quadruped robot constructed from the LocoKit modular robot. The learning strategy applies a stochastic optimization algorithm to optimize eight open parameters of a central pattern generator based gait implementation. We observe that the strategy converges in roughly ten minutes to gaits of similar or higher velocity than a manually designed gait and that the strategy readapts in the event of failed actuators. In future work we plan to study co-learning of morphological and control parameters directly on the physical robot. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Platform for building large-scale agent-based systems

    Page(s): 69 - 73
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (602 KB) |  | HTML iconHTML  

    This paper presents an agent platform called PANGEA (Platform for Automatic construction of orGanizations of intElligent Agents). This platform allows to developed multiagent systems modeled as Virtual Organizations. The concepts of roles, organizations and norms are fully supported by the platform assuring flexibility and scalability. Moreover, a communication protocol based on IRC gives high performance and reliability to this kind of distributed systems. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • On path planning: Adaptation to the environment using Fast Marching

    Page(s): 74 - 79
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1783 KB) |  | HTML iconHTML  

    This paper presents a novel methodology for fast path planning based on an offline predefined skeleton of the environment by means of the Fast Marching Square method. The FM2 skeleton concept is introduced whose intuition is a set which contains the shortest paths (in terms of time) of a given environment This way, the path planning algorithm is adapted to the environment where a robot will navigate. An algorithm to obtain this skeleton is detailed in order to be able to use it for path planning purposes. When planning paths over the skeleton, the results show that the time reduction is about 50% while the characteristics of the FM2 method remain: completeness, safeness, smoothness and absence of local minima. This paper also shows that the proposed method performs well in more than 2 dimensions. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A kinodynamic planning-learning algorithm for complex robot motor control

    Page(s): 80 - 83
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1114 KB) |  | HTML iconHTML  

    Robot motor control learning is currently one of the most active research areas in robotics. Many learning techniques have been developed for relatively simple problems. However, very few of them have direct applicability in complex robotic systems without assuming prior knowledge about the task, mainly due to three facts. Firstly, they scale badly to continues and high dimensional problems. Secondly, they need too many real robot-environment interactions. Finally, they are not capable of adapting to environment or robot dynamic changes. In order to overcome these problems, we have developed a new algorithm capable of finding from scratch open-loop state-action trajectory solutions by mixing sample-based tree kinodynamic planning with dynamic model learning. Some results demonstrating the viability of this new type of approach in the cart-pole swing-up task problem are presented. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An adaptive energy-aware routing protocol for MANETs using the SARSA reinforcement learning algorithm

    Page(s): 84 - 89
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1050 KB) |  | HTML iconHTML  

    In MANETs (Mobile Ad-hoc NETworks), communicating nodes are powered by batteries which could not be re-charged in many practical usage scenarios. Hence, maximizing network lifetime is a critical optimization objective in routing protocols design for MANETs. To meet this objective, energy-consumption should be balanced among all mobile nodes. In this paper, we formulate the energy-aware route discovery problem in a reactive routing protocol as a Reinforcement Learning (RL) problem that we solve using the SARSA RL algorithm. We have implemented our proposed RL-model on the top of AODV a well-known reactive routing protocol for MANETs. Furthermore, we show through simulations the efficiency of our proposal, against an implementation of the Energy-Aware Probability routing protocol. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Autonomous visual self-localization in completely unknown environment

    Page(s): 90 - 95
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1269 KB) |  | HTML iconHTML  

    In this paper, a novel approach to visual self-localization in an unknown environment is presented. The proposed method makes possible the recognition of new landmark without using GPS or any other communication links or pre-training. An image-based self-localization technique is used to automatically label landmarks that are detected in real-time using a computationally efficient and recursive algorithm. Real-time experiments are carried in outdoor environment at Lancaster University using a real mobile robot Pioneer 3DX in order to build a map the local environment without using any communication links. The presented experimental results in real situations show the effectiveness of the proposed method. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.