H.3. Artificial Intelligence
Amir Mehrabinezhad; Mohammad Teshnelab; Arash Sharifi
Abstract
Due to the growing number of data-driven approaches, especially in artificial intelligence and machine learning, extracting appropriate information from the gathered data with the best performance is a remarkable challenge. The other important aspect of this issue is storage costs. The principal component ...
Read More
Due to the growing number of data-driven approaches, especially in artificial intelligence and machine learning, extracting appropriate information from the gathered data with the best performance is a remarkable challenge. The other important aspect of this issue is storage costs. The principal component analysis (PCA) and autoencoders (AEs) are samples of the typical feature extraction methods in data science and machine learning that are widely used in various approaches. The current work integrates the advantages of AEs and PCA for presenting an online supervised feature extraction selection method. Accordingly, the desired labels for the final model are involved in the feature extraction procedure and embedded in the PCA method as well. Also, stacking the nonlinear autoencoder layers with the PCA algorithm eliminated the kernel selection of the traditional kernel PCA methods. Besides the performance improvement proved by the experimental results, the main advantage of the proposed method is that, in contrast with the traditional PCA approaches, the model has no requirement for all samples to feature extraction. As regards the previous works, the proposed method can outperform the other state-of-the-art ones in terms of accuracy and authenticity for feature extraction.
Y. Sharafi; M. Teshnelab; M. Ahmadieh Khanesar
Abstract
A new multi-objective evolutionary optimization algorithm is presented based on the competitive optimization algorithm (COOA) to solve multi-objective optimization problems (MOPs). Based on nature-inspired competition, the competitive optimization algorithm acts between animals such as birds, cats, bees, ...
Read More
A new multi-objective evolutionary optimization algorithm is presented based on the competitive optimization algorithm (COOA) to solve multi-objective optimization problems (MOPs). Based on nature-inspired competition, the competitive optimization algorithm acts between animals such as birds, cats, bees, ants, etc. The present study entails main contributions as follows: First, a novel method is presented to prune the external archive and at the same time keep the diversity of the Pareto front (PF). Second, a hybrid approach of powerful mechanisms such as opposition-based learning and chaotic maps is used to maintain the diversity in the search space of the initial population. Third, a novel method is provided to transform a multi-objective optimization problem into a single-objective optimization problem. A comparison of the result of the simulation for the proposed algorithm was made with some well-known optimization algorithms. The comparisons show that the proposed approach can be a better candidate to solve MOPs.
H. Sadr; Mir M. Pedram; M. Teshnehlab
Abstract
With the rapid development of textual information on the web, sentiment analysis is changing to an essential analytic tool rather than an academic endeavor and numerous studies have been carried out in recent years to address this issue. By the emergence of deep learning, deep neural networks have attracted ...
Read More
With the rapid development of textual information on the web, sentiment analysis is changing to an essential analytic tool rather than an academic endeavor and numerous studies have been carried out in recent years to address this issue. By the emergence of deep learning, deep neural networks have attracted a lot of attention and become mainstream in this field. Despite the remarkable success of deep learning models for sentiment analysis of text, they are in the early steps of development and their potential is yet to be fully explored. Convolutional neural network is one of the deep learning methods that has been surpassed for sentiment analysis but is confronted with some limitations. Firstly, convolutional neural network requires a large number of training data. Secondly, it assumes that all words in a sentence have an equal contribution to the polarity of a sentence. To fill these lacunas, a convolutional neural network equipped with the attention mechanism is proposed in this paper which not only takes advantage of the attention mechanism but also utilizes transfer learning to boost the performance of sentiment analysis. According to the empirical results, our proposed model achieved comparable or even better classification accuracy than the state-of-the-art methods.
Gh. Ahmadi; M. Teshnelab
Abstract
Because of the existing interactions among the variables of a multiple input-multiple output (MIMO) nonlinear system, its identification is a difficult task, particularly in the presence of uncertainties. Cement rotary kiln (CRK) is a MIMO nonlinear system in the cement factory with a complicated mechanism ...
Read More
Because of the existing interactions among the variables of a multiple input-multiple output (MIMO) nonlinear system, its identification is a difficult task, particularly in the presence of uncertainties. Cement rotary kiln (CRK) is a MIMO nonlinear system in the cement factory with a complicated mechanism and uncertain disturbances. The identification of CRK is very important for different purposes such as prediction, fault detection, and control. In the previous works, CRK was identified after decomposing it into several multiple input-single output (MISO) systems. In this paper, for the first time, the rough-neural network (R-NN) is utilized for the identification of CRK without the usage of MISO structures. R-NN is a neural structure designed on the base of rough set theory for dealing with the uncertainty and vagueness. In addition, a stochastic gradient descent learning algorithm is proposed for training the R-NNs. The simulation results show the effectiveness of proposed methodology.
H.6.2.2. Fuzzy set
N. Moradkhani; M. Teshnehlab
Abstract
Cement rotary kiln is the main part of cement production process that have always attracted many researchers’ attention. But this complex nonlinear system has not been modeled efficiently which can make an appropriate performance specially in noisy condition. In this paper Takagi-Sugeno neuro-fuzzy ...
Read More
Cement rotary kiln is the main part of cement production process that have always attracted many researchers’ attention. But this complex nonlinear system has not been modeled efficiently which can make an appropriate performance specially in noisy condition. In this paper Takagi-Sugeno neuro-fuzzy system (TSNFS) is used for identification of cement rotary kiln, and gradient descent (GD) algorithm is applied for tuning the parameters of antecedent and consequent parts of fuzzy rules. In addition, the optimal inputs of the system are selected by genetic algorithm (GA) to achieve less complexity in fuzzy system. The data related to Saveh White Cement (SWC) factory is used in simulations. The Results demonstrate that the proposed identifier has a better performance in comparison with neural and fuzzy models have presented earlier for the same data. Furthermore, in this paper TSNFS is evaluated in noisy condition which had not been worked out before in related researches. Simulations show that this model has a proper performance in different noisy condition.
F.2.7. Optimization
M. Kosari; M. Teshnehlab
Abstract
Although many mathematicians have searched on the fractional calculus since many years ago, but its application in engineering, especially in modeling and control, does not have many antecedents. Since there are much freedom in choosing the order of differentiator and integrator in fractional calculus, ...
Read More
Although many mathematicians have searched on the fractional calculus since many years ago, but its application in engineering, especially in modeling and control, does not have many antecedents. Since there are much freedom in choosing the order of differentiator and integrator in fractional calculus, it is possible to model the physical systems accurately. This paper deals with time-domain identification fractional-order chaotic systems where conventional derivation is replaced by a fractional one with the help of a non-integer derivation. This operator is itself approximated by a N-dimensional system composed of an integrator and a phase-lead filter. A hybrid particle swarm optimization (PSO) and genetic algorithm (GA) method has been applied to estimate the parameters of approximated nonlinear fractional-order chaotic system that modeled by a state-space representation. The feasibility of this approach is demonstrated through identifying the parameters of approximated fractional-order Lorenz chaotic system. The performance of the proposed algorithm is compared with the genetic algorithm (GA) and standard particle swarm optimization (SPSO) in terms of parameter accuracy and cost function. To evaluate the identification accuracy, the time-domain output error is designed as the fitness function for parameter optimization. Simulation results show that the proposed method is more successful than other algorithms for parameter identification of fractional order chaotic systems.
A.1. General
F. Alibakhshi; M. Teshnehlab; M. Alibakhshi; M. Mansouri
Abstract
The stability of learning rate in neural network identifiers and controllers is one of the challenging issues which attracts great interest from researchers of neural networks. This paper suggests adaptive gradient descent algorithm with stable learning laws for modified dynamic neural network (MDNN) ...
Read More
The stability of learning rate in neural network identifiers and controllers is one of the challenging issues which attracts great interest from researchers of neural networks. This paper suggests adaptive gradient descent algorithm with stable learning laws for modified dynamic neural network (MDNN) and studies the stability of this algorithm. Also, stable learning algorithm for parameters of MDNN is proposed. By proposed method, some constraints are obtained for learning rate. Lyapunov stability theory is applied to study the stability of the proposed algorithm. The Lyapunov stability theory is guaranteed the stability of the learning algorithm. In the proposed method, the learning rate can be calculated online and will provide an adaptive learning rare for the MDNN structure. Simulation results are given to validate the results.
