Sharif Digital Repository

In this paper we have proposed a multi-agent architecture for an Intelligent Tutoring System using TROPOS methodology. Basic assumptions of constructivist learning have been taken into account in analysis and design activities. Nine types of agents are distinguished in the presented system. We have centralized management in the proposed architecture which is done by an agent called Manager Agent. This strategy of centralized management will result in a more coherent and effective Tutoring System than other systems not using such an agent, since it applies control flow of agents and decreases agents ' useless communications which may cause destructive congestion in system. This text discusses... 

Graph Matching (GM) plays an essential role in computer vision and machine learning. The ability of using pairwise agreement in GM makes it a powerful approach in feature matching. In this paper, a new formulation is proposed which is more robust when it faces with outlier points. We add weights to the one-to-one constraints, and modify them in the process of optimization in order to diminish the effect of outlier points in the matching procedure. We execute our proposed method on different real and synthetic databases to show both robustness and accuracy in contrast to several conventional GM methods  

In this paper, we address the problem of dictionary learning for sparse representation. Considering the regularized form of the dictionary learning problem, we propose a method based on a homotopy approach, in which the regularization parameter is overall decreased along iterations. We estimate the value of the regularization parameter adaptively at each iteration based on the current value of the dictionary and the sparse coefficients, such that it preserves both sparse coefficients and dictionary optimality conditions. This value is, then, gradually decreased for the next iteration to follow a homotopy method. The results show that our method has faster implementation compared to recent... 

The MAXQ hierarchical reinforcement learning method is computationally expensive in applications with deep hierarchy. In this paper, we propose a new learning algorithm for MAXQ method to address the open problem of reducing its computational complexity. While the computational cost of the algorithm is considerably decreased, the required storage of new algorithm is less than two times as the original learning algorithm requires storage. Our experimental results in the simple Taxi Domain Problem show satisfactory behavior of the new algorithm  

Taxonomy learning is one of the major steps in ontology learning process. Manual construction of taxonomies is a time-consuming and cumbersome task. Recently many researchers have focused on automatic taxonomy learning, but still quality of generated taxonomies is not satisfactory. In this paper we have proposed a new compound similarity measure. This measure is based on both knowledge poor and knowledge rich approaches to find word similarity. We also used Neural Network model for combination of several similarity methods. We have compared our method with simple syntactic similarity measure. Our measure considerably improves the precision and recall of automatic generated taxonomies. ©... 

In this work we assume that there is an agent in an unknown environment (domain). This agent has some predefined actions and it can perceive its current state in the environment completely. The mission of this agent is to fulfill the tasks (goals) that are often assigned to it as fast as it can. Acting has lots of cost, and usually planning and simulating the environment can reduce this cost. In this paper we address a new approach for incremental induction of probabilistic planning operators, from this environment while the agent tries to reach to its current goals. It should be noted that there have been some works related to incremental induction of deterministic planning operators and... 

Web-based E-Leaning Systems (WELSs) have emerged as new means of skill training and knowledge acquisition, encouraging both academia and industry to invest resources in the adoption of these systems. Traditionally, most pre- and post-adoption tasks related to evaluation are carried out from the viewpoints of technology. Since users have been widely recognized as being a key group of stakeholders in influencing the adoption of information systems, their attitudes about these systems are considered as pivotal. Therefore, based on the theory of multi-criteria decision making and the research results concerning user satisfaction in the fields of human-computer interaction and information... 

Kernel learning is one of the most important and recent approaches to constrained clustering. Until now many kernel learning methods have been introduced for clustering when side information in the form of pairwise constraints is available. However, almost all of the existing methods either learn a whole kernel matrix or learn a limited number of parameters. Although the non-parametric methods that learn whole kernel matrix can provide capability of finding clusters of arbitrary structures, they are very computationally expensive and these methods are feasible only on small data sets. In this paper, we propose a kernel learning method that shows flexibility in the number of variables between... 

In this paper an approach for blind source separation in bilinear (or linear quadratic) mixtures is presented. The proposed algorithm employs the same recurrent structure as [Hosseini and Deville, 2003) for separating these mixtures . However, instead of maximal likelihood, our algorithm is based on minimizing the mutual information of the outputs for recovering the independent components. Simulation results show the efficiency of the proposed algorithm. © 2009 IEEE  

Most email users have encountered with spam problems, which have been addressed as a text classification or categorization problem. In this paper, we propose a novel spam detection method that uses ensemble of classifiers based on clustering and selection techniques. There is diversity in genre of e-mail's content and this method can find different topics in emails by clustering. It first computes disjoint clusters of emails, and then a classifier is trained on each cluster. When new email arrives, its cluster is identified. The classifier of the identified cluster is selected to classify the new email. Our method can extract many kinds of topics in emails. The evaluation shows that the... 

In this paper, an adaptive random search method, based on continuous action-set learning automata, is studied for solving stochastic optimization problems in which only the noisecorrupted value of a function at any chosen point in the parameter space is available. First, a new continuous action-set learning automaton is introduced and its convergence properties are studied. Then, applications of this new continuous action-set learning automata to the minimization of a penalized Shubert function and pattern classification are presented. © Sharif University of Technology  

Zamin artificial life model is designed to be a general purpose environment for researches on evolution of learning methods, living strategies and complex behaviors and is used in several studies thus far. As a main target for Zamin's design has been its expandability and ease of problem definition, a new agent based structure for this artificial world is introduced in this paper, which is believed to be much easier to use and extend. In this new model, all control and world running processes are done by agents. Therefore, any change in world processes does not require recoding the main engine and can be done just by altering the behavior of one or some agents. To have an easier interface... 

In this paper we propose a fast and efficient algorithm for learning overcomplete dictionaries. The proposed algorithm is indeed an alternative to the well-known K-Singular Value Decomposition (K-SVD) algorithm. The main drawback of K-SVD is its high computational load especially in high-dimensional problems. This is due to the fact that in the dictionary update stage of this algorithm an SVD is performed to update each column of the dictionary. Our proposed algorithm avoids performing SVD and instead uses a special form of alternating minimization. In this way, as our simulations on both synthetic and real data show, our algorithm outperforms K-SVD in both computational load and the quality... 

While recent supervised dictionary learning methods have attained promising results on the classification tasks, their performance depends on the availability of the large labeled datasets. However, in many real world applications, accessing to sufficient labeled data may be expensive and/or time consuming, but its relatively easy to acquire a large amount of unlabeled data. In this paper, we propose a probabilistic framework for discriminative dictionary learning which uses both the labeled and unlabeled data. Experimental results demonstrate that the performance of the proposed method is significantly better than the state of the art dictionary based classification methods  

In this work we present a sparse dictionary learning method, specifically tuned to solve the tracking problem. Recently, sparse representation has drawn much attention because of its genuineness and strong mathematical background. In this paper we present an online method for dictionary learning which is desirable for problems such as tracking. Online learning methods are preferable because the whole data are not available at the current time. The presented method tries to use the advantages of the generative and discriminative models to achieve better performance. The experimental results show our method can overcome many tracking challenges  

In this paper, we introduce a new hybrid learning algorithm, called DTGP, to construct cost-sensitive classifiers. This algorithm uses a decision tree as its basic classifier and the constructed decision tree will be pruned by a genetic programming algorithm using a fitness function that is sensitive to misclassification costs. The proposed learning algorithm has been examined through six cost-sensitive problems. The experimental results show that the proposed learning algorithm outperforms in comparison to some other known learning algorithms like C4.5 or naïve Bayesian  

This paper proposes four methods for improving the performance of keyword spotting (KWS) systems. Keyword models are usually created by concatenating the phoneme HMMs and garbage models consist of all phonemes HMMs. We present the results of investigations involving the use of skips in states of keyword HMMs and we focus on improving the hit ratio; then for false alarm reduction in KWS we model the words that are similar to keywords and we create HMMs for highly frequent words. These models help to improve the performance of the filler model. Two post-processing steps based on phoneme and word probabilities are used on the results of KWS to reduce the false alarms. We evaluate the... 

Scattered and little research in the field of Persian speech synthesis systems has been performed during the last ten years. Comprehensive framework that properly implements and adapts statistical speech synthesis methods for Persian has not been conducted yet. In this paper, recent statistical parametric speech synthesis methods including CLUSTERGEN, traditional HMM-based speech synthesis and its STRAIGHT version, are implemented and adapted for Persian language. CCR test is carried out to compare these methods with each other and with unit selection method. Listeners Score samples based on CMOS. The methods were ranked by averaging the CCR scores. The results show that STRAIGHT-based... 

In this study a new method is proposed for quantification of cardiac muscle motions in echocardiography frames based on empirical mode decomposition (EMD) and manifold learning method. EMD algorithm is able to extract intrinsic mode functions (IMF) from a signal. In the first bi-dimension intrinsic mode functions (BIMF) of echocardiography frames myocardial is shown with more details than the second BIMF and the second BIMF shows more details than the third BIMF. By using manifold learning method, quantification difference between each pair of consecutive frames in the first, second and third BIMF series (similarities between the frames were extracted). Acquired trajectories of three... 

This study aimed to examine the effect of robot assisted language learning (RALL) on the anxiety level and attitude in English vocabulary acquisition amongst Iranian EFL junior high school students. Forty-six female students, who were beginners at the age of 12, participated in this study and were randomly assigned into two groups of RALL (30 students) and non-RALL (16 students). The textbook, the materials, as well as the teacher were the same in the two groups. However in the RALL group, the treatment was given by a teacher accompanied by a humanoid robot assistant. Two questionnaires of anxiety and attitude were utilized to measure the students’ anxiety and attitude (Horwitz et al. 1986;...