Optimal Local Basis: A Reinforcement Learning Approach for Face Recognition

This paper presents a novel learning approach for Face Recognition by introducing Optimal Local Basis. Optimal local bases are a set of basis derived by reinforcement learning to represent the face space locally. The reinforcement signal is designed to be correlated to the recognition accuracy. The optimal local bases are derived then by finding the most discriminant features for different parts of the face space, which represents either different individuals or different expressions, orientations, poses, illuminations, and other variants of the same individual. Therefore, unlike most of the existing approaches that solve the recognition problem by using a single basis for all individuals, our proposed method benefits from local information by incorporating different bases for its decision. We also introduce a novel classification scheme that uses reinforcement signal to build a similarity measure in a non-metric space. Experiments on AR, PIE, ORL and YALE databases indicate that the proposed method facilitates robust face recognition under pose, illumination and expression variations. The performance of our method is compared with that of Eigenface, Fisherface, Subclass Discriminant Analysis, and Random Subspace LDA methods as well.     

Robot learning language — Integrating programming and learning for cognitive systems

One central property of cognitive systems is the ability to learn and to improve continually. We present a robot control language that combines programming and learning in order to make learning executable in the normal robot program. The language constructs of our learning language RoLL rely on the concept of hierarchical hybrid automata to enable a declarative, explicit specification of learning problems. Using the example of an autonomous household robot, we point out some instances where learning–and especially continued learning–makes the robot control program more cognitive.     

Adaptive binary tree for fast SVM multiclass classification

This paper presents an adaptive binary tree (ABT) to reduce the test computational complexity of multiclass support vector machine (SVM). It achieves a fast classification by: (1) reducing the number of binary SVMs for one classification by using separating planes of some binary SVMs to discriminate other binary problems; (2) selecting the binary SVMs with the fewest average number of support vectors (SVs). The average number of SVs is proposed to denote the computational complexity to exclude one class. Compared with five well-known methods, experiments on many benchmark data sets demonstrate our method can speed up the test phase while remain the high accuracy of SVMs.     

Applying a novel decision rule to the sphere-structured support vector machines algorithm

The traditional sphere-structured support vector machines algorithm is one of the learning methods. It can partition the training samples space by means of constructing the spheres with the minimum volume covering all training samples of each pattern class in high-dimensional feature space. However, the decision rule of the traditional sphere-structured support vector machines cannot assign ambiguous sample points such as some encircled by more than two spheres to valid class labels. Therefore, the traditional sphere-structured support vector machines is insufficient for obtaining the better classification performance. In this article, we propose a novel decision rule applied to the traditional sphere-structured support vector machines. This new decision rule significantly improves the performance of labeling ambiguous points. Experimental results of seven real datasets show the traditional sphere-structured support vector machines based on this new decision rule can not only acquire the better classification accuracies than the traditional sphere-structured support vector machines but also achieve the comparable performance to the classical support vector machines. An erratum to this article can be found at     

Optimal assembly plan generation: a simplifying approach

The main difficulty in the overall process of optimal assembly plan generation is the great number of different ways to assemble a product (typically thousands of solutions). This problem confines the application of most existing automated planning methods to products composed of only a limited number of components. The presented method of assembly plan generation belongs to the approach called “disassembly” and is founded on a new representation of the assembly process, with introduction of a new concept, the equivalence of binary trees. This representation allows to generate the minimal list of all non-redundant (really different) assembly plans. Plan generation is directed by assembly operation constraints and plan-level performance criteria. The method was tested for various assembly applications and compared to other generation approaches. Results show a great reduction in the combinatorial explosion of the number of plans. Therefore, this simplifying approach of assembly sequence modeling allows to handle more complex products with a large number of parts.     

An adaptive genetic-based signature learning system for intrusion detection

Rule-based intrusion detection systems generally rely on hand crafted signatures developed by domain experts. This could lead to a delay in updating the signature bases and potentially compromising the security of protected systems. In this paper, we present a biologically-inspired computational approach to dynamically and adaptively learn signatures for network intrusion detection using a supervised learning classifier system. The classifier is an online and incremental parallel production rule-based system.A signature extraction system is developed that adaptively extracts signatures to the knowledge base as they are discovered by the classifier. The signature extraction algorithm is augmented by introducing new generalisation operators that minimise overlap and conflict between signatures. Mechanisms are provided to adapt main algorithm parameters to deal with online noisy and imbalanced class data. Our approach is hybrid in that signatures for both intrusive and normal behaviours are learnt.The performance of the developed systems is evaluated with a publicly available intrusion detection dataset and results are presented that show the effectiveness of the proposed system.     

Symbolic hierarchical analysis in currency markets: An application to contagion in currency crises

In this paper we introduce a new method to describe dynamical patterns of the real exchange rate co-movements time series and to analyze contagion in currency crisis. The method combines the tools of symbolic time series analysis with the nearest neighbor single linkage clustering algorithm. Data symbolization allows us obtaining a metric distance between two different time series that is used to construct an ultrametric distance. By analyzing the data of various countries, we derive a hierarchical organization, constructing minimal-spanning and hierarchical trees. From these trees we detect different clusters of countries according to their proximity. We show that this methodology permits us to construct a structural and dynamic topology that is useful to study interdependence and contagion effects among financial time series.     

Least squares twin support vector machines for pattern classification

In this paper we formulate a least squares version of the recently proposed twin support vector machine (TSVM) for binary classification. This formulation leads to extremely simple and fast algorithm for generating binary classifiers based on two non-parallel hyperplanes. Here we attempt to solve two modified primal problems of TSVM, instead of two dual problems usually solved. We show that the solution of the two modified primal problems reduces to solving just two systems of linear equations as opposed to solving two quadratic programming problems along with two systems of linear equations in TSVM. Classification using nonlinear kernel also leads to systems of linear equations. Our experiments on publicly available datasets indicate that the proposed least squares TSVM has comparable classification accuracy to that of TSVM but with considerably lesser computational time. Since linear least squares TSVM can easily handle large datasets, we further went on to investigate its efficiency for text categorization applications. Computational results demonstrate the effectiveness of the proposed method over linear proximal SVM on all the text corpuses considered.     

双核模式的组播过渡系统的设计与实现

IPv6作为下一代互联网的核心技术,必然会逐渐取代IPv4.但在相当长的一段时间内,IPv6会与IPv4共存.在对以往一些组播过渡技术进行研究的基础上,通过linux的内核编程来实现一个基于双核模式的组播过渡系统.将该系统部署在IPv6域与IPv4域的边界,能实现两个域之间的组播通信.同时,该组播过渡系统只有在需要进行转换时才开始工作,从而提高了转换器的工作效率.