Face recognition is one of the most important and challenging issues in computer vision and image processing. About half a century ago, since the first face recognition system was introduced, facial recognition has become one of the most important issues in industry and academia. In recent years, with the developing of computers throughput and developments of a new generation of hierarchical learning algorithms called deep learning, much attention has been devoted to solving learning problems by deep learning algorithms. Deep neural networks perform feature learning instead of feature extraction which by this strategy they are much useful for image processing and computer vision problems. Deep neural network through feature learning perform data representation well and have gained many successes in learning and complex problems, many studies have been done on the application of deep neural networks to face recognition and many successes has been achieved. In this study we examine the neural network based methods used for face recognition such as multilayer perceptrons, restricted Boltzmann machine and auto encoders. Most of our study devoted to convolutional neural network as one of the most successful deep learning algorithms. At the end we have examined the results of the encountered methods on ORL, AR, YALE, FERET datasets and show deep neural network has gained high recognition rate in comparing with benchmark methods.
Meta‑Heuristic algorithms are optimization techniques that provide the optimal solution through processes of repeated exploration and exploitation of the entire search space. Feature selection is also an important and prominent process in the field of machine learning that reduces data dimensions. This paper examines and compares nature-inspired meta-heuristic algorithms for feature selection to increase the accuracy of software fault prediction. Researchers cannot easily select meta-heuristic algorithms as a suitable method for their research due to their great variety and multiplicity. In this paper, by describing the feature selection techniques and its methods, the application of meta-heuristic algorithms in different fields, such as swarm intelligence and binary methods of these algorithms has been investigated. Also, by introducing 18 meta-heuristic algorithms in 6 different categories and evaluating each of them, a suitable analysis has been provided to researchers so that they can easily and with the highest efficiency choose the appropriate algorithm and method of their work. In the papers presented so far, meta-heuristic algorithms have been studied from only one aspect, while in this article, while studying different types of research, they have tried to study and evaluate them from different aspects. The effectiveness of the combination of three meta-heuristic algorithms, developed butterfly flame, bee colony and developed wall, was tested on 20 data sets. the proposed method in 17 datasets was able to improve the result of 7 datasets.
This paper proposes an evolutionary computing method called quantum buzzard optimization algorithm (QBUZO). The QBUZO algorithm is a modification of the BUZO algorithm based on quantum system. In quantum, the particle can be represented by its instantaneous motion and the energy of its wave function. In QBUZO, each particle is defined with quantum state in this system and is established with a wave function instead of the position and capability vector in BUZO algorithm. The idea is to allow all particles to have quantum behavior instead of the classical Newtonian random motion. We present the individual particle of a BUZO system moving in quantum multidimensional space, and create a quantum delta potential well model for BUZO. The effectiveness of QBUZO is demonstrated by experimental results on some benchmark mathematical functions in optimization. The comparison of performances with six other optimization algorithms represents that QBUZO is superior to other algorithms.
1 January 2025 in International Journal of Engineering
The ever-increasing growth of amoral content in cyberspace has caused the idea of detecting such contents in images, then creating an alarm or limiters in different age ranges around the world. This content not only detects harmful visual items, but also it detects instances of abusing individuals. This study has represented a new approach to detect images with amoral content. The suggested model includes two deep models based on inception blocks and attention and residual based convolutional layer which acts with two different approaches: the first model is a five-class recognition model and the second one is an estimator model which maps the image into levels of inappropriateness between 0 and 1. Combining these two models by an aggregator based on first-order rules leads to representing a model which improved 2.1% detection precision in not suitable for work (NSFW) dataset and 2.3% on TI-UNRAM dataset in comparison with other state-of-the-art models. Our approach also shows promising results to model culturally based definition of NSFW, especially in Iran.