In making the deep neural network, activation functions play an important role. But the choice of activation functions also affects the network in term of optimization and to retrieve the better results. Several activation functions have been introduced in machine learning for many practical applications. But which activation function should use at hidden layer of deep neural networks was not identified. The primary objective of this analysis was to describe which activation function must be used at hidden layers for deep neural networks to solve complex non-linear problems. The configuration for this comparative model was used by using the datasets of 2 classes (Cat/Dog). The number of Convolutional layer used in this network was 3 and the pooling layer was also introduced after each layer of CNN layer. The total of the dataset was divided into the two parts. The first 8000 images were mainly used for training the network and the next 2000 images were used for testing the network. The experimental comparison was done by analyzing the network by taking different activation functions on each layer of CNN network. The validation error and accuracy on Cat/Dog dataset were analyzed using activation functions (ReLU, Tanh, Selu, PRelu, Elu) at number of hidden layers. Overall the Relu gave best performance with the validation loss at 25th Epoch 0.3912 and validation accuracy at 25th Epoch 0.8320. It is found that a CNN model with ReLU hidden layers (3 hidden layers here) gives best results and improve overall performance better in term of accuracy and speed. These advantages of ReLU in CNN at number of hidden layers are helpful to effectively and fast retrieval of images from the databases.
Image fusion is to converge Multispectral (MS) and Panchromatic (PAN) images into a fused image which is further enlightened. Soft computing based image fusion techniques are fuzzy and neuro-fuzzy are exploited to lessening the severance and vagueness in the output. Fused images achieved after the synthesis is utilized in image analysis, medical applications, armed province, and computer revelation. In this research, we convey iterative image fusion based on fuzzy and neuro-fuzzy methods on source images attained from different sources to improve visualization proficiency. We also compared the proposed techniques with principal component analysis (PCA) and wavelets transform based image fusion. Fused outcomes accomplished from image fusion methods are assessed through typical eminence evaluation parameters. The resulting outcome obtained from iterative fusion is improved in terms of spectral and spatial information when compared to the one-time fused image. Due to neural networks structure, applied sorts of biological neural networks and potentiality of the fuzzy and neuro-fuzzy logic, the proposed method overtakes the conventional methods. The complete investigational consequences formed from anticipated methodology established that the utilization of proposed approaches enhanced image content.
Background: A spiral phase rear mounted masked scheme is proposed based on Gyrator Transform (GT) to enhance the security contribution of second lens of the existing Double Random Phase Encoding (DRPE) system by modulating the phase of the output obtained in output plane. An additional third layer of Spiral Phase Mask (SPM) is included in the output plane in the same 4f system. Objective: To develop a symmetric cryptosystems to enhance the security potential of the second lens and to prevent the comfortable realization of the cipher-image in the transform domain. Methods: The original image is first scrambled using Arnold transform with frequency and then is convoluted with a secret random phase mask, RPM in GT and then obtained result is convoluted with another secret RPM in inverse GT. The obtained result is then finally convoluted with SPM. Results: It verifies the sensitivity and achieves better performance in terms of recovering a high quality image. Results show the security, performance and quality analysis on the basis of correlation coefficient, occlusion attack, key sensitivity and noise attack, entropy and histogram. Conclusion: It enhances the security potential of second lens in DRPE and introduces diffusion in the system. The system is simulated for binary and greyscale image and achieves better performance as compared to existing DRPE variants. Key sensitivity is more secure and cannot recover original image without knowing all the parameters. Correlation coefficient are also weakly correlated and does not reveals relevant information. Simulation result demonstrates the feasibility and robustness of cryptosystem.
Purpose: The aim of this article is to discuss the impact of static load balancing over a set of heterogeneous processors, where tasks are independent and unitary in static environments, by showing how to distribute task in order to optimize both the average response time and the degree of the resources used. Methods: Implementation of a modified scheduling algorithm, the latter is based on two parameters which are the execution time and the failure probability. The algorithm is based on the results of an optimal algorithm that already exists, with only one parameter that is execution time. Results: The obtained results show that the modified scheduling algorithm gives us the good results. Conclusion: The modified algorithm assumes that the processor has smallest execute time. So, the failure probability increases because of it’s frequently use. The results obtained by testing this proposed algorithm are better than the optimal algorithm.
Background: The whale optimization algorithm is one of the popular meta-heuristic algorithms which has successfully been applied in various application areas such as image analysis and data clustering. However, the slow convergence rate and chances of sticking into the local optima due to improper balance of its exploration and exploitation phases are some of its pitfalls. Therefore, in this paper, a new improved whale optimization algorithm has been proposed. Moreover, the proposed method has been used in bag-of-features method for histopathological image classification. Methods: The new algorithm, improved whale optimization algorithm, modifies the encircling phase of original whale optimization algorithm. The proposed algorithm has been used to cluster the extracted features for finding the relevant codewords to be used in the bag-of-features method for histopathological image classification. Results: The efficiency of proposed algorithm has been analyzed on 23 benchmark functions in terms of mean fitness, standard deviation values, and convergence behavior. The performance of the improved whale optimization algorithm based histopathological image classification method has been analyzed on blue histology image dataset and compared with other meta-heuristic based bagof- features methods in terms of recall, precision, F-measure, and accuracy. The experimental results validate that the proposed method outperforms the considered state-of-the-art methods and attains 12% increase in the histopathological image classification accuracy. Conclusion: In this paper, a new improved whale optimization algorithm has been proposed and applied in bag-of-features method for histopathological image classification. The results of proposed method outperform the other existing meta-heuristic methods over standard benchmark functions and histopathological image dataset.
