Sharif Digital Repository

Incremental learning has been used extensively for data stream classification. Most attention on the data stream classification paid on non-evolutionary methods. In this paper, we introduce new incremental learning algorithms based on harmony search. We first propose a new classification algorithm for the classification of batch data called harmony-based classifier and then give its incremental version for classification of data streams called incremental harmony-based classif ier. Finally, we improve it to reduce its computational overhead in absence of drifts and increase its robustness in presence of noise. This improved version is called improved incremental harmony-based classifier. The... 

Autoregressive models are among the most successful neural network methods for estimating a distribution from a set of samples. However, these models, such as other neural methods, need large data sets to provide good estimations. We believe that knowing structural information about the data can improve their performance on small data sets. Masked autoencoder for distribution estimation (MADE) is a well-structured density estimator, which alters a simple autoencoder by setting a set of masks on its connections to satisfy the autoregressive condition. Nevertheless, this model does not benefit from extra information that we might know about the structure of the data. This information can... 

Autoregressive models are among the most successful neural network methods for estimating a distribution from a set of samples. However, these models, such as other neural methods, need large data sets to provide good estimations. We believe that knowing structural information about the data can improve their performance on small data sets. Masked autoencoder for distribution estimation (MADE) is a well-structured density estimator, which alters a simple autoencoder by setting a set of masks on its connections to satisfy the autoregressive condition. Nevertheless, this model does not benefit from extra information that we might know about the structure of the data. This information can... 

Expert finding problem in bibliographic networks has received increased interest in recent years. This problem concerns finding relevant researchers for a given topic. Motivated by the observation that rarely do all coauthors contribute to a paper equally, in this paper, we propose two discriminative methods for realizing leading authors contributing in a scientific publication. Specifically, we cast the problem of expert finding in a bibliographic network to find leading experts in a research group, which is easier to solve. We recognize three feature groups that can discriminate relevant experts from other authors of a document. Experimental results on a real dataset, and a synthetic one... 

Cartesian Genetic Programming (CGP) has an unusual genotype representation which makes it more efficient than Genetic programming (GP) in digital circuit design problem. However, to the best of our knowledge, all methods used in evolutionary design of digital circuits deal with rugged, complex search space, which results in long running time to obtain successful evolution. Therefore, employing a method to guide evolution in these spaces can facilitate achieving more reasonable results. It has been claimed that a two-step evolutionary scenario caused by benefit and cost of learning called Baldwin effect can guide evolution in the biology and artificial life. Therefore, we have been motivated... 

Nowadays developing autonomous systems, which can act in various environments and interactively perform their assigned tasks, are intensively desirable. These systems would be ready to be applied in different fields such as medicine, controller robots and social life. Reinforcement learning is an attractive area of machine learning which addresses these concerns. In large scales, learning performance of an agent can be improved by using hierarchical Reinforcement Learning techniques and temporary extended actions. The higher level of abstraction helps the learning agent approach lifelong learning goals. In this paper a new method is presented for discovering subgoal states and constructing... 

Cellular learning automata (CLA) is a distributed computational model which was introduced in the last decade. This model combines the computational power of the cellular automata with the learning power of the learning automata. Cellular learning automata is composed from a lattice of cells working together to accomplish their computational task; in which each cell is equipped with some learning automata. Wide range of applications utilizes CLA such as image processing, wireless networks, evolutionary computation and cellular networks. However, the only input to this model is a reinforcement signal and so it cannot receive another input such as the state of the environment. In this paper,... 

Density peaks clustering (DPC) is as an efficient clustering algorithm due for using a non-iterative process. However, DPC and most of its improvements suffer from the following shortcomings: (1) highly sensitive to its cutoff distance parameter, (2) ignoring the local structure of data in computing local densities, (3) using a crisp kernel to calculate local densities, and (4) suffering from the cause of chain reaction. To address these issues, in this paper a new method called DPC-DBFN is proposed. The proposed method uses a fuzzy kernel for improving separability of clusters and reducing the impact of outliers. DPC-DBFN uses a density-based kNN graph for labeling backbones. This strategy... 

Community Question Answering is one of the valuable information resources which provide users with a platform to share their knowledge. Finding potential experts in CQA is beneficial to several problems like low participation rate of the users, long waiting time to receive answers and to the low quality of answers. Many research papers focused on retrieving the expert users of CQAs. Most of them are taking expertise into consideration at the query time and ignore the temporal aspects of the expert finding problem. However, considering the evolution of personal expertise over time can improve the quality of expert finding. In many applications, it is beneficial to find the potential experts... 

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  

An stochastic version of the classical shortest path problem whereby for each node of a graph, a probability distribution over the set of successor nodes must be chosen so as to reach a certain destination node with minimum expected cost. In this paper, we propose a new algorithm based on Particle Swarm Optimization (PSO) for solving Stochastic Shortest Path Problem (SSPP). The comparison of our algorithm with other algorithms indicates that its performance is suitable even by the less number of iterations. © 2008 Springer-Verlag Berlin Heidelberg  

Automatic image annotation is a mechanism to assign a list of appropriate tags that describe the visual content of a given image. Most methods only focus on the content of the images and ignore the relationship between the tags in vocabulary. In this work, we propose a new deep learning-based automatic image annotation architecture, which considers label dependencies in a graph convolution neural network structure and extracts tag descriptors to re-weight the output class scores based on their relationships. The proposed architecture has three main parts: feature extraction, graph convolutional network, and annotation. In graph convolutional network, we apply one layer convolution on... 

[No abstract available]  

Image segmentation is one of the fundamental problems in image processing and computer vision, since it is the first step in many image analysis systems. This paper presents a new perspective to image segmentation, namely, segmenting input images by applying efficient community detection algorithms common in social and complex networks. First, a common segmentation algorithm is used to fragment the image into small initial regions. A weighted network is then constructed. Each initial region is mapped to a vertex, and all these vertices are connected to each other. The similarity between two regions is calculated from colour information. This similarity is then used to assign weights to the... 

Data streams are endless flow of data produced in high speed, large size and usually non-stationary environments. The main property of these streams is the occurrence of concept drifts. Using decision trees is shown to be a powerful approach for accurate and fast learning of data streams. In this paper, we present an incremental regression tree that can predict the target variable of newly incoming instances. The tree is updated in the case of occurring concept drifts either by altering its structure or updating its embedded models. Experimental results show the effectiveness of our algorithm in speed and accuracy aspects in comparison to the best state-of-the-art methods  

Handling recurring concepts has become of interest as a challenging problem in the field of data stream classification in recent years. One main feature of data streams is that they appear in nonstationary environments. This means that the concept which the data are drawn from, changes over the time. If after a long enough time, the concept reverts to one of the previous concepts, it is said that recurring concepts has occurred. One solution to this challenge is to maintain a pool of classifiers, each representing a concept in the stream. This paper follows this approach and holds an ensemble of classifiers for each concept. As for each received batch of data, a new classifier is created;... 

Data streams have some unique properties which make them applicable in precise modeling of many real data mining applications. The most challenging property of data streams is the occurrence of "concept drift". Recurring concepts is a type of concept drift which can be seen in most of real world problems. Detecting recurring concepts makes it possible to exploit previous knowledge obtained in the learning process. This leads to quick adaptation of the learner whenever a concept reappears. In this paper, we propose a learning algorithm called Pool and Accuracy based Stream Classification with some variations, which takes the advantage of maintaining a pool of classifiers to track recurring... 

The goal of image segmentation is to find regions that represent objects or meaningful parts of objects. In this paper a new method is presented for color image segmentation which involves the ideas used for community detection in social networks. In the proposed method an initial segmentation is applied to partition input image into small homogeneous regions. Then a weighted network is constructed from the regions, and a community detection algorithm is applied to it. The detected communities represent segments of the image. A remarkable feature of the method is the ability to segments the image automatically by optimizing the modularity value in the constructed network. The performance of... 

One of the main challenges of data streams is the occurrence of concept drift. Concept drift is the change of target (or feature) distribution, and can occur in different types: sudden, gradual, incremental or recurring. Because of the forgetting mechanism existing in the data stream learning process, recurring concepts has received much attention recently, and became a challenging problem. This paper tries to exploit the existence of recurring concepts in the learning process and improve the classification of data streams. It uses a pool of concepts to detect the reoccurrence of a concept using two methods: a Bayesian, and a heuristic method. Two approaches are used in the classification... 

In this paper, a novel reinforcement learning method inspired by the way humans learn from others is presented. This method is developed based on cellular learning automata featuring a modular design and cooperation techniques. The modular design brings flexibility, reusability and applicability in a wide range of problems to the method. This paper focuses on analyzing sensitivity of the method's parameters and the applicability in optimization problems. Results of the experiments justify that the new method outperforms similar ones because of employing knowledge sharing technique, reasonable exploration logic, and learning rules based on the action trajectory