Sharif Digital Repository

With the increasing volume of data, the best approach for learning from this data is to exploit an online learning algorithm. Online ensemble methods take advantage of an ensemble of classifiers to predict labels of data. Prediction with expert advice is a well-studied problem in the online ensemble learning literature. The weighted majority and the randomized weighted majority (RWM) algorithms are two well-known solutions to this problem, aiming to converge to the best expert. Since among some expert, the best one does not necessarily have the minimum error in all regions of data space, defining specific regions and converging to the best expert in each of these regions will lead to a... 

A dictionary learning algorithm learns a set of atoms from some training signals in such a way that each signal can be approximated as a linear combination of only a few atoms. Most dictionary learning algorithms use a two-stage iterative procedure. The first stage is to sparsely approximate the training signals over the current dictionary. The second stage is the update of the dictionary. In this paper we develop some atom-by-atom dictionary learning algorithms, which update the atoms sequentially. Specifically, we propose an efficient alternative to the well-known K-SVD algorithm, and show by various experiments that the proposed algorithm is much faster than K-SVD while its results are... 

During the last decade, there has been a growing interest toward the problem of sparse decomposition. A very important task in this field is dictionary learning, which is designing a suitable dictionary that can sparsely represent a group of training signals. In most dictionary learning algorithms, the cost function to determine the the optimum dictionary is the ℓ0 norm of the matrix of decomposition coefficients of the training signals. However, we believe that this cost function fails to fully express the goal of dictionary learning, because it only sparsifies the whole set of coefficients for all training signals, rather than the coefficients for each training signal individually. Thus,... 

Free gait becomes necessary in walking robots when they come to walk over discontinuous terrain or face some difficulties in walking. A basic gait generation strategy is presented here using reinforcement learning and fuzzy reward approach. A six-legged (hexapod) robot is implemented using Q-learning algorithm. The learning ability of walking in a hexapod robot is explored considering only the ability of moving its legs and using a fuzzy rewarding system telling whether and how it is moving forward. Results show that the hexapod robot learns to walk using the presented approach properly  

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  

In this paper the problem of automatically discovering subtasks and hierarchies in reinforcement learning is considered. We present a novel method that allows an agent to autonomously discover subgoals and create a hierarchy from actions. Our method identifies subgoals by partitioning local state transition graphs. Options constructed for reaching these subgoals are added to action choices and used for accelerating the Q-Learning algorithm. Experimental results show significant performance improvements, especially in the initial learning phase  

We explore the dictionary learning problem for sparse representation when the signals are dependent. In this paper, a first-order Markovian model is considered for dependency of the signals, that has many applications especially in medical signals. It is shown that the considered dependency among the signals can degrade the performance of the existing dictionary learning algorithms. Hence, we propose a method using the Maximum Log-likelihood Estimator (MLE) and the Expectation Minimization (EM) algorithm to learn the dictionary from the signals generated under the first-order Markovian model. Simulation results show the efficiency of the proposed method in comparison with the... 

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  

Several reasons such as no free lunch theorem indicate that there is not a universal Feature selection (FS) technique that outperforms other ones. Moreover, some approaches such as using synthetic dataset, in presence of large number of FS techniques, are very tedious and time consuming task. In this study to tackle the issue of dependency of estimation accuracy on the selected FS technique, a methodology based on the heterogeneous ensemble is proposed. The performance of the major learning algorithms of neural network (i.e. the FFNN-BR, the FFNN-LM) in combination with the diverse FS techniques (i.e. the NCA, the F-test, the Kendall's tau, the Pearson, the Spearman, and the Relief) and... 

Demand forecasting has always been a concern for business owners as one of the main activities in supply chain management. Unlike the past, that forecasting was done with the help of a limited amount of information, today, using advanced technologies and data analytics, forecasting is performed with machine learning algorithms and data-driven methods. Patterns and trends of demand, customer information, preferences, suggestions, and post-consumption feedbacks are some types of data that are used in various demand forecasting efforts. Traditional statistical methods and techniques are biased in demand prediction and are not accurate; so, machine learning algorithms as more popular techniques... 

A new algorithm for learning jointly reconstructive and discriminative dictionaries for sparse representation (SR) is presented. While in a usual dictionary learning algorithm like K-SVD only the reconstructive aspect of the sparse representations is considered to learn a dictionary, in our proposed algorithm, which we call K-LDA, the discriminative aspect of the sparse representations is also addressed. In fact, K-LDA is an extension of K-SVD in the case that the class informations (labels) of the training data are also available. K-LDA takes into account these information in order to make the sparse representations more discriminate. It makes a trade-off between the amount of... 

In this study, the artificial neural network (ANN) was used for the prediction of WDT. The inputs to network are molar mass and pressure, and the output is WDT at each input. A two-layer network with different hidden neurons and different learning algorithms such as LM, SCG, GDA and BR were examined. The network with 16 hidden neurons and Levenberg-Marquardt (LM) train function showed the best results in comparison with the other networks. Also, the predicted results of this network were compared with the thermodynamic models and better accordance with experimental data for ANN was concluded  

Dimensionality reduction is a necessary task in data mining when working with high dimensional data. A type of dimensionality reduction is feature selection. Feature selection based on feature ranking has received much attention by researchers. The major reasons are its scalability, ease of use, and fast computation. Feature ranking methods can be divided into different categories and may use different measures for ranking features. Recently, ensemble methods have entered in the field of ranking and achieved more accuracy among others. Accordingly, in this paper a Heterogeneous ensemble based algorithm for feature ranking is proposed. The base ranking methods in this ensemble structure are... 

During recent years, active learning has evolved into a popular paradigm for utilizing user's feedback to improve accuracy of learning algorithms. Active learning works by selecting the most informative sample among unlabeled data and querying the label of that point from user. Many different methods such as uncertainty sampling and minimum risk sampling have been utilized to select the most informative sample in active learning. Although many active learning algorithms have been proposed so far, most of them work with binary or multi-class classification problems and therefore can not be applied to problems in which only samples from one class as well as a set of unlabeled data are... 

This paper deals with the problem of multi-agent learning of a population of players, engaged in a repeated normal-form game. Assuming boundedly-rational agents, we propose a model of social learning based on trial and error, called "social reinforcement learning". This extension of well-known Q-learning algorithm, allows players within a population to communicate and share their experiences with each other. To illustrate the effectiveness of the proposed learning algorithm, a number of simulations on the benchmark game of "Battle of Sexes" has been carried out. Results show that supplementing communication to the classical form of Q-learning, significantly improves convergence speed towards... 

This letter proposes a dictionary learning algorithm for blind one bit compressed sensing. In the blind one bit compressed sensing framework, the original signal to be reconstructed from one bit linear random measurements is sparse in an unknown domain. In this context, the multiplication of measurement matrix A and sparse domain matrix φ, i.e., D = Aφ, should be learned. Hence, we use dictionary learning to train this matrix. Towards that end, an appropriate continuous convex cost function is suggested for one bit compressed sensing and a simple steepest-descent method is exploited to learn the rows of the matrix D. Experimental results show the effectiveness of the proposed algorithm... 

This letter proposes a dictionary learning algorithm for solving the grid mismatch problem in direction of arrival (DOA) estimation from both the array sensor data and from the sign of the array sensor data. Discretization of the grid in the sparsity-based DOA estimation algorithms is a problem, which leads to a bias error. To compensate this bias error, a dictionary learning technique is suggested, which is based on minimizing a suitable cost function. We also propose an algorithm for the estimation of DOA from the sign of the measurements. It extends the iterative method with adaptive thresholding algorithm to the 1-b compressed sensing framework. Simulation results show the effectiveness... 

Several reasons such as no free lunch theorem indicates that any learning algorithm in combination with a specific feature selection (FS) technique may give more accurate estimation than other learning algorithms. Therefore, there is not a universal approach that outperforms other algorithms. Moreover, due to the large number of FS techniques, some recommended solutions such as using synthetic dataset or combining different FS techniques are very tedious and time consuming. In this study to tackle the issue of more accurate estimation of NPPs parameters, the performance of the major supervised learning algorithms in combination with the different FS techniques which are appropriate for... 

This paper develops a novel insight and procedure that includes a variety of algorithms for finding the best solution in a structured multi-agent system with internal communications and a global purpose. In other words, it finds the optimal communication structure among agents and the optimal policy in this structure. First, a unique reinforcement learning algorithm is proposed to find the optimal policy of each agent in a fixed structure with non-linear function approximators like artificial neural networks (ANN) and with eligibility traces. Secondly, a mechanism is presented to perform self-organization based on the information of the learned policy. Finally, an algorithm that can discover...