Abstract: Artificial neural networks (ANNs) have shown efficient results in statistics and computer science applications. Feed-forward neural network (FNN) is the most popular and simplest neural network architecture, capable of solving nonlinearity. In this paper, feed-forward neural networks’ weight and bias figuring using a newly proposed metaheuristic Salp Swarm Algorithm (SSA) are proposed. SSA is a swarm-based metaheuristic inspired by the navigating and foraging behaviour of salp swarm. The performance is evaluated for some of the benchmarked datasets and compared with some well-known metaheuristics.

Abstract: The main aim of any path/motion planning algorithm in the context of a mobile robot is to produce a collision-/crash-free path among the goal and start points in an environment in which it is present. The past few decades have seen the development of various methodologies to design an optimal path. The present research focuses on the development of an optimized path planning algorithm for the robot using a hybrid method after combining particle swarm optimization (PSO) algorithm with potential field method for static obstacles and potential field method (PFM) prediction for dynamic obstacles. While implementing, PSO-based potential field method, the total potential, that is the sum of repulsive and attractive potentials, is considered as the fitness function which is optimized using PSO algorithm. Further, a 3-point method has been used for smoothing the obtained path. Once the image of the scene is obtained, a clustering method is employed to find the center of obstacle and the location of the robot has been determined by calculating the repulsive potential in each iteration. Finally, the developed algorithms are tested on both the static and dynamic environments in computer simulations and found satisfactory.

Abstract: With this tremendous growth of the Internet, mobile devices, and e-business, information load is increasing day by day. That leads to the development of the system, which can filter and prioritize the relevant information for users. Recommendation system solves this issue by enabling users to get knowledge, products, and services of personalized basis. Since the inception of recommender system, researcher has paid much attention and developed various filtering techniques to make these systems effective and efficient in terms of users and system experience. This paper presents a preliminary survey of different recommendation system based on filtering techniques, challenges applications, and evaluation metrics. The motive of work is to introduce researchers and practitioner with the different characteristics and possible filtering techniques of recommendation systems.

Abstract: Most of the real-world optimization problems have multiple objectives to deal with. Satisfying one objective at a time may lead to the huge deviation in other. Therefore, an efficient tool is required which can handle multiple objectives simultaneously in order to provide a set of desired solutions. In view of this, multi-objective optimization (MOO) attracts the attention of the researchers since last few decades. Many classical optimization techniques have been proposed by the researchers to solve the multi-objective optimization problems. However mostly, the gradient-based approaches fail to handle complex MOO problems. Hence, as an alternative, researchers have shown their interest toward population-based optimization approaches to solve the MOO problems and come up with convincing results even in the complex environment. Evolutionary algorithms (EAs), which are the first in the group of population-based approach, enjoy almost a decade in providing the solutions to MOO problems. The real challenge is to achieve the set of solutions called Pareto-optimal set. The smooth landing on such set is only possible if there exists diversified solution in the population. Due to the continuous effort, there is a gradual development in the proposition of various efficient Pareto-based approaches in the literature to solve MOEAs. A critical review of those approaches is being carried out in this present study.

Abstract: This paper presents dynamic performance comparison of a fuzzy logic-based proportional, integral, and derivative controller (FPID) with different membership functions such as triangular, trapezoidal, and Gaussian for load frequency control (LFC) in an interconnected two-area thermal power system. The parameters of controller are optimized by using spider monkey optimization (SMO) algorithm. The superiority of the proposed algorithm is established by comparing the results with popularly used algorithms like particle swarm optimization (PSO) and teaching–learning-based optimization (TLBO). Initially, the linearized model of the system is considered with reheat turbine; then, the study is extended by imposing nonlinearity such as generation rate constraints (GRC) and governor dead band (GDB). The result comparison is analyzed using various time domain specifications like peak undershoot, peak overshoot, and settling time of different area frequencies and tie-line power deviation between them applying a step load perturbation (SLP) of 1%.

Abstract: Grey Wolf Optimizer (GWO), developed by Mirjalili et al. (Adv Eng Softw 69:46–61, 2014 [1]), is a recently developed nature-inspired technique based on leadership hierarchy of grey wolves. In this paper, Grey Wolf Optimizer has been hybridized with differential evolution (DE) mutation, and two versions, namely DE-GWO and gDE-GWO, have been proposed to avoid the stagnation of the solution. To evaluate the performance of both the proposed versions, a set of 23 well-known benchmark problems has been taken. The comparison of obtained results between original GWO and proposed hybridized versions of GWO is done with the help of Wilcoxon signed-rank test. The results conclude that the proposed hybridized version gDE-GWO of GWO has better potential to solve these benchmark test problems compared to GWO and DE-GWO.

Abstract: The medical imaging field has its significance with an increase in the demand for automatic and efficient diagnosis in a brief time period. In this paper, the Fuzzy C-means (FCM) algorithm is used to identify the tumor and extract it and parameters like segmented area, Mean Squared Error (MSE) and Peak Signal to Noise Ratio (PSNR) are found. The daubechies three level decomposition of Discrete 2-D Wavelet Transform (DWT) is used to extract the coefficients of wavelets for the Magnetic Resonance (MR) image and then its dimensionality is reduced by Principle Component Analysis (PCA) algorithm. Gray level co-occurrence matrix of these coefficients are found and then thirteen statistical features are extracted from the given MR image. The extracted features describe the input image’s texture and the structural information of the intensity. These extracted features of the training set of images constitute the training feature database. This database is used for training and a test input image is given to Support Vector Machine (SVM) classifier to classify 105 MR brain tumor images into either benign or malignant class. From the obtained experimental results, the proposed SVM classifier had an accuracy of 98.82%, 100% of sensitivity, 97.83% of specificity, and 1.17% of error rate.

Abstract: Human gender is an important demographic characteristic in the society. Recognizing demography characteristics of individuals, for example, age and gender using automatic image recognition taken much consideration in last few years. This paper proposes the extraction of geometric and appearance feature of face automatically from the front view. For extracting the feature, cumulative benchmark approach is used. Two basic categories as supervised as well as unsupervised methodology may be applied for gender grouping. In this paper, we used supervised machine learning approach. We have used three diverse classifiers, for this approach as SVM, neural network, and adobos. We have trained all the classifiers by means of identical training dataset and similar feature. We have done a comparative study of the performance of these classifiers and which classifier is best for our primary dataset over face images.

Abstract: Vehicles play a vital role in modern-day transportation systems. Number plate provides a standard means of identification for any vehicle. To serve this purpose, automatic license plate recognition system was developed. This consisted of four major steps: preprocessing of obtained image, extraction of license plate region, segmentation, and character recognition. In earlier research, direct application of Sobel edge detection algorithm or applying threshold was used as key steps to extract the license plate region, which do not produce efficient results when captured image is subjected to high intensity of light. The use of morphological operations causes deformity in the characters during segmentation. We propose a novel algorithm to tackle the mentioned issues through a unique edge detection algorithm. It is also a tedious task to create and update the database of required vehicles frequently. This problem is solved by the use of ‘Internet of things’ where an online database can be created and updated from any module instantly. Also, through IoT, we connect all the cameras in a geographical area to one server to create a ‘universal eye’ which drastically increases the probability of tracing a vehicle over having manual database attached to each camera for identification purpose.

Abstract: This paper presents a deep learning approach for ear localization and recognition. The comparable complexity between human outer ear and face in terms of its uniqueness and permanence has increased interest in the use of ear as a biometric. But similar to face recognition, it poses challenges such as illumination, contrast, rotation, scale, and pose variation. Most of the techniques used for ear biometric authentication are based on traditional image processing techniques or handcrafted ensemble features. Owing to extensive work in the field of computer vision using convolutional neural networks (CNNs) and histogram of oriented gradients (HOG), the feasibility of deep neural networks (DNNs) in the field of ear biometrics has been explored in this research paper. A framework for ear localization and recognition is proposed that aims to reduce the pipeline for a biometric recognition system. The proposed framework uses HOG with support vector machines (SVMs) for ear localization and CNN for ear recognition. CNNs combine feature extraction and ear recognition tasks into one network with an aim to resolve issues such as variations in illumination, contrast, rotation, scale, and pose. The feasibility of the proposed technique has been evaluated on USTB III database. This work demonstrates 97.9% average recognition accuracy using CNNs without any image preprocessing, which shows that the proposed approach is promising in the field of biometric recognition.

Abstract: Handwriting recognition is an important application of pattern recognition subject. Although some research studies of handwriting recognition of a few major Indian scripts can be found in the literature, the same is not true for many of the Indian scripts. Malayalam is one such script, and automatic recognition issues of this script remain largely unexplored till date. On the other hand, there are nearly 40 million people mainly living in the southern part of India whose native language is Malayalam. In the present article, we present our recent study of Malayalam offline handwritten word recognition. The main contributions of the present study are (a) pioneering development of a database for offline handwritten word samples of Malayalam script and (b) its benchmark recognition results based on a transfer learning strategy which involves a deep convolutional neural network (CNN) architecture for feature extraction and a support vector machine (SVM) for classification. Recognition result of the proposed architecture on the writer independent test set of Malayalam handwritten word sample database is quite satisfactory. Moreover, the same architecture has been found to improve the existing state of the art of offline handwriting recognition of several major Indian scripts.

Abstract: The expanding textual information and significance of examining the substance has started a colossal research in the field of synopsis. Text summarization is the process of conveying the gist of a text with a minimized representation. The requirement for automation of the procedure is at its apex with exponential burst of information because of digitization. Text captioning comes under the branch of abstractive summarization which captures the gist of the article in a few words. In this paper, we present an approach to text captioning using recurrent neural networks which comprise of an encoder–decoder model. The key challenges dealt here was to figure out the ideal input required to produce the desired output. The model performs better when the input is fed with the summary as compared to the original article itself. The recurrent neural network model with LSTM results has been effective in transcribing a caption for the textual data.

Abstract: This paper proposes an improved sine–cosine algorithm for solving power distribution network reconfiguration (PDNR) problem. The sine–cosine algorithm is a recently proposed population-based meta-heuristic optimization algorithm which uses the mathematical sine and cosine functions for searching the solution space. The search procedure looks for the best solution by repeatedly making small changes to an initial solution until no further improved solutions are found. To maintain a balance between local and global search, four random variables (r1, r2, r3 and r4) are integrated into this algorithm. For applying this algorithm to the PDNR problem, some improvements are proposed in this meta-heuristic search algorithm along with a new data structure-based load flow method to minimize power loss as the single objective. The effectiveness of the proposed PDNR algorithm is tested by considering five standard test distribution systems (33, 69, 84, 119 and 136 buses).

Abstract: This paper presents a method for path testing by generating the test data automatically and optimizing the test data to test the critical paths for a software under test (SUT), using real-coded genetic algorithm. Real encoding is used for automatic test data generation, and a representative test suite, which achieves 100% path coverage, is found as an optimum result. In this paper, the proposed real-coded genetic algorithm for path coverage (RCGAPC) generates a set of inputs for testing a specific software and outperforms by giving effective and efficient results in terms of less number of test data generation counts. In the proposed approach, one-to-one injective mapping scheme is used for mapping the test data to the corresponding path and the most critical path is covered during path testing of a specific software. It seems to be faster than the traditional GA in covering critical path. The proposed method can reduce the number of test data generation required for path testing of a SUT and give an optimized Test suite that covers 100% path for specific software.

Abstract: This paper investigates the applicability of artificial neural network model for strength prediction of fibers’ self-compacting concrete under compression. The available 99 experimental data samples of fibers self-compacting concrete were used in this research work. In this paper, computational-based research is carried for predicting the strength of concrete under compression and model was developed using ANN with five input nodes and feed-forward three-layer back-propagation neural networks with ten hidden nodes were examined using learning algorithm. ANN model proposed analytically was verified, and it gives more compatible results. Hence, the ANN model is proposed to predict the strength of fibrous self-compacting concrete under compression.

Abstract: Transmission congestion is a vital problem in a deregulated power system. This paper proposes a novel transmission congestion management approach considering photovoltaic (PV) power using lion optimization algorithm (LOA). The main contributions of this paper have twofolds. Initially, the values of bus sensitivity factor (BSF) and generator sensitivity factor (GSF) are, respectively, used to select the optimal bus to integrate PV power and to select the participating generators for congestion management. Finally, LOA is used to determine the active power rescheduling amount and congestion cost. Test results on modified 39 bus New England system indicate that the LOA approach could provide a less active power rescheduling amount and congestion cost with integration of PV power compared to particle swarm optimization (PSO) and ant lion optimizer (ALO) algorithm.

Abstract: In this modern era, one of the prime most facilities available to this generation is state-of-the-art health care, and still diabetes has emerged as one the leading chronic disease. Diabetes is a condition which implies the glucose level is more than the inquisitive level on a managed premise. The prime motto of this study is to provide a good classification of diabetes. There are existing methods, which are for the classification of diabetes popularly datasets “Pima Indian Diabetes Dataset.” Here, the proposed work comprises of four phases: In the first stage, a “Localized Diabetes Dataset” has been compiled and collected from Bombay Medical Hall, Upper Bazar Ranchi, India. In the second stage, neural networks has been used as the classification technique on localized diabetes dataset. In the third stage, GA has been used as a feature selection technique through which six features among twelve features have been obtained. Lastly in the fourth stage, neural networks have been used for classification on suitable attributes produced by GA. In this study, the results have been compared with and without GA for used classification technique. It has been concluded in this work that GA is helpful in removing not only significant attributes, deducing the cost and computation time but also enhancing the ROC and accuracy. The utilized strategy may likewise be executed in other medical issues.

Abstract: Mutation testing is a fault-based unit testing in which faults are detected by executing certain test data designed by any white box testing technique. This paper presents a hybridized method for path testing as well as mutation testing by generating the test data automatically using genetic algorithm. In the proposed approach, first path coverage-based test data is generated and further this data is exercised to cover all mutants present in the specific program under test. The proposed method can improve the testing efficiency by deleting the redundant test data obtained from the path testing in terms of better mutation score, and fault detection matrix is used to delete the duplicate data covering same mutants.

Abstract: This paper presents the coordinated navigation problem of multiple wheeled robots. Two motion planners such as PFM and GA-tuned FLC are combined with a coordination strategy block to solve such problems. Performances of the developed approaches are tested through computer simulations. A total hundred numbers of scenarios are taken to show the efficacy of the proposed navigation schemes. GA-tuned FLC has wholly outperformed the PFM in most of the situations. Also, with the increase in some robots, coordination count has increased, and the need for the strategies was prominent. It is experienced that proposed coordination strategy along with the developed motion planners results in a time-optimal and realistic solution to the discussed problem.

Abstract: In recent days, enormous data are generated from departmental stores, hospitals, social media, banks, etc. These datasets are associated with different association rules for monitoring the business operations. During this process, to avoid leaking of sensitive information leads to development of association rule hiding algorithms. Many heuristic algorithms are developed but they are limited to optimal solutions. In this paper, an efficient meta-heuristic algorithm has been developed for association rule hiding based on chemical reaction optimization algorithm. The results of the proposed approach are compared with the genetic algorithm, particle swarm optimization, and cuckoo-based algorithms. The experimental results of the proposed algorithm are tested on the benchmark datasets.

Abstract: Cloud computing besides being used in industries it is also used in academics; existing cloud computing architectures do not support cloud forensic investigations and are also not forensic ready to a remarkable extent, and also the present tools which are being used in the cloud forensics do not support the elastic nature of cloud. We explore and expose several issues related to cloud forensics in cloud computing by keeping an eye on the concepts of cloud computing which are being developed and are utilized along with latest technologies and also the investments which are being made on cloud computing. Latest developments in technologies have created certain challenges which are emerging and have exposed that cloud has the potential to handle most computing technologies which are being transformative, one such challenging concepts which have been increasing its wait is cloud forensics. In this paper we have traced out the concepts which revolves around cloud forensics and here we propose a model which includes a VM SnapShot Server which continuously stores the snapshots of the cloud service provider and certain servers involved in computing moment by moment so that this would be useful for any digital crime related to cloud, as this plays a key role in identifying the correct cause of the mischief task which resulted in the loss or damage of the original data; this is also helpful during the cases where either the cloud service provider or the suspect gives an incorrect information during the investigation carried out in digital crime; this model also has certain advantages over the present existing models. When certain new activities such as uploading a malware in the cloud, downloading more files then the permissible number, more access from a location, cracking the saved passwords, launching and deleting malicious files, creating corrupted files on the sensitive data stored in the cloud such kind of things can also be traced out easily.

Abstract: Image retrieval is an integral part of many different search engines. Search based on metadata of the image has been a primary approach in the process of image retrieval. In this work, we implement a search engine for better quality image retrieval using query image. Our implementation uses elastic search for indexing of the available images in the server and intermediate captioning mechanism for both search and retrieval process. The image captioning has been carried out using VGG16 Convolutional Neural Network. The implemented engine has been implemented and tested using the popular benchmark dataset called Flickr-8k dataset. The retrieved image quality demonstrated promising performance and suggests that an intermediate captioning-based image search could be an alternative to metadata-based search engines.

Abstract: Question answering (QA) is a field of Natural Language Processing that deals with generating answers automatically to questions asked to a system. It can be categorized into two types—open-domain and closed-domain QA. Open-domain QA can deal with questions about anything, whereas closed-domain QA deals with questions in a specific domain. In our work, we use the architectures of LSTM and memory networks to perform closed-domain question answering and compare the performances of the two. LSTMs are specialized RNNs that can remember necessary data and forget the irrelevant bits. Since data in QA consist of stories and questions based on them, this model seems appropriate, with the ability to handle long sequences. On the other hand, memory networks provide an architecture where there is a provision to store the information learnt by the system in an explicit memory component, rather than just as weight matrices. This also seems like an architecture well-suited to question answering. We implement each model and train it on the Facebook bAbi dataset. This dataset is specifically generated for the purpose of evaluating QA systems on the twenty prerequisite toy bAbi tasks. Each dataset corresponds to one task and checks whether the model is able to perform chaining, counting, answer with single and multiple supporting facts, understand relations, directions, etc. Based on the performances of each model on the bAbi tasks, we perform a comparative study of the two.

Abstract: This chapter presents the experiment of symmetrical cascaded switched-diode multilevel inverter (SCSD MLI) with fuzzy controller for a different number of levels. The objective of this topology is to reduce the number of power semiconductor switches along with its gate driver circuits as the number of level increases; therefore, the complexity of the circuit and installation cost of the converter are reduced when compared with the conventional cascaded multilevel inverter and cascaded half-bridge multilevel inverter. In this document, seven-, nine-, eleven-level SCSD MLIs are analyzed with simulation results which describe the total harmonic distortion reduction with the increment in number of levels. For this circuit topology, phase disposition pulse-width modulation technique is developed to regulate the RMS output voltage of inverter. In order to maintain the RMS output voltage, appropriate fuzzy controllers are constructed. MATLAB/SIMULINK simulation results of seven-level, nine-level, and eleven-level are presented to justify the performance of the suggested topology.

Abstract: Single Image Super-Resolution techniques have the function of retrieving a high resolution image from a single low resolution input. They implement deep learning heuristics which perform the techniques to form pixel-accourate reproductions. In this paper we have experimented upon various neural architectures with unique approaches towards the task of super-resolution. We have especially elaborated upon adversarial training networks which are yielding progressive results in both conditional and quantifiable benchmarks.

Abstract: This paper introduces a two-stage clustering model to determine the seismic activities of a region in spatio-temporal domain. In the spatial domain for cluster analysis, a K-means algorithm based on “Haversine distance” is introduced. With this, a seismic region is classified into distinct zones which are correlated in space. In each zone, events’ temporal activities are analyzed. This temporal domain analysis is carried out using a variable “\(\varepsilon \)” density-based clustering algorithm. In this algorithm, the neighborhood radius “\(\varepsilon \)” is varied to determine the core points. The variation of “\(\varepsilon \)” is a time-dependent function of “magnitude” of the event (empirical relation makes out higher magnitude leads to a larger value of “\(\varepsilon \)”, i.e., number of days in time). The proposed model is applied to analyze the seismic activities of Himalaya and Sumatra–Andaman region for the time interval between 1965 and 2015 (51 Years). Simulation results reveal that the de-clustered catalogs obtained for both the regions follow linear trend which justifies the background events are homogeneous with respect to time. Corresponding clustered catalogs which reflect the presence of foreshock, mainshock, and aftershock events follow the behavior of true catalog with time.

Abstract: Image enhancement is one of the key concerns pertaining to better quality image photography captured through modern digital cameras. Probability of digital images getting compromised through lightning and weather conditions remains high. Due to these environmental limitations, many a time loss of information from images is reported. Major role of image amplification is to bring out hidden details of an image from the sample. It provides multiple options for enhancing the visual quality of images. This paper addresses problem of early skin cancer detection using image enhancement techniques and presents a multi-scale retinex with color restoration (MSR-CR) technique for skin cancer detection. The actual skin portion suffering from cancer is identified by comparing enhanced image with available ground truth image. Experimental result shows significant improvement over previously available techniques.

Abstract: In recent years, hyperspectral image (HSI) classification has become a popular topic of the research. The common problem with HSI is imbalance between limited number of available samples and high dimensionality. To deal with this issue, several linear and nonlinear feature reduction approaches can be used. In HSI, another issue is the imbalance in number of labeled samples present in different classes. This paper presents a novel approach for dealing with imbalanced learning problem in HSI. In this proposed approach, first principal component analysis (PCA) algorithm is applied for feature reduction, followed by application of synthetic minority oversampling technique (SMOTE) on the reduced dimensional dataset. In order to estimate the percentage of oversampling required for each class, the particle swarm optimization technique (PSO) is used. After wisely oversampling the samples present in each class, the oversampled training data is fed into the k-nearest neighbor (KNN) classifier. The obtained results revealed that by properly oversampling the training samples per class, the classification accuracy is increased with reduced time complexity. The proposed approach was tested on the widely used Indian Pines dataset.

Abstract: An experimental investigation has been performed on wire-EDM of Inconel 718 by using zinc-coated brass wire as tool electrode. Based on L27 orthogonal array, the studies have been conducted by varying five process parameters (such as taper angle, pulse on-time, wire speed, wire tension, and discharge current), within the selected experimental domain. The following machining performance yields (viz. angular error and surface roughness) have been investigated during this study. The angular error (AE) and surface roughness (Ra) have been critical responses in die making industry. Simultaneously, the mathematical model has been developed by using nonlinear regression analysis for correlating with various process parameters and performance characteristics. Finally, JAYA algorithm has been applied individually to find out the satisfactory machining performances. Application and potential of JAYA algorithm have been compared with Teaching-Learning based algorithm and Genetic Algorithm (GA) and observed by the JAYA algorithm, which does not require any specific parameter settings and hence easy to implement.