In this study, a modified class topper optimization algorithm named as MCTO is proposed for clustering problem. In the MCTO, cloning, crossover and mutation operators are introduced for global and faster convergence a...
In this study, a modified class topper optimization algorithm named as MCTO is proposed for clustering problem. In the MCTO, cloning, crossover and mutation operators are introduced for global and faster convergence and better quality solutions. Two different types of hybridizations with K-means named as K-MCTO and MCTO-K are then performed with the MCTO to check the effectiveness. First, local optimal solutions obtained from the K-means is fed to the MCTO to get a global optimal solution. Then, another hybridization is done by feeding the solutions obtained from the MCTO to the K-means. Experimental study shows that the MCTO, K-MCTO and MCTO-K outperform various well-known state-of-the-arts. The algorithms are validated by clustering five standard data sets taken from agriculture, health-care and geology domains.
The COVID-19 pandemic has increased demand for face mask detection systems that utilize deep learning and machine learning algorithms. However, these systems are susceptible to adversarial attacks, where an attacker c...
详细信息
The COVID-19 pandemic has increased demand for face mask detection systems that utilize deep learning and machine learning algorithms. However, these systems are susceptible to adversarial attacks, where an attacker can manipulate the system to make incorrect predictions. This study aimed to test the vulnerability of a deep learning-based face mask detection model to a specific type of attack called a black box adversarial attack in which the attacker possesses only partial information about the target model. The study's findings showed that the attack successfully reduced the model's accuracy from 96.48% to 49.25%. This emphasizes the need for more robust defense mechanisms in face mask detection systems to ensure their reliability.
Accurate analysis of Side Scan Sonar (SSS) images has potential applications in the identification of submerged objects like ships, planes etc. The underwater SSS images are inherently contaminated with noise. In this...
Accurate analysis of Side Scan Sonar (SSS) images has potential applications in the identification of submerged objects like ships, planes etc. The underwater SSS images are inherently contaminated with noise. In this manuscript, a bilateral filtering algorithm is applied for noise reduction. After the noise reduction, the objective is to segment the image and determine the object, shadow, and background. The image segmentation is dealt here as a multi-objective clustering problem. Minimization of intra-cluster distance and negative of inter-cluster distance are taken as the two objective functions for analysis. The optimization task is carried out by four benchmark multi-objective optimization algorithms SMS-EMOA, RVEA, MOEA/D, and NSGA-II. Simulation study on four KLSG SSS Images reveals that the performance obtained by the SMS-EMOA algorithm is superior compared to its counterpart algorithms.
Malware is intended to harm computers or networks, and it frequently entails engaging in unlawful or forbidden activity that can be utilised for espionage or financial gain. Malware assaults are now beginning to affec...
Malware is intended to harm computers or networks, and it frequently entails engaging in unlawful or forbidden activity that can be utilised for espionage or financial gain. Malware assaults are now beginning to affect embedded computational platforms, including Internet of Things (IoT) devices, medical equipment, and environmental and industrial control systems. This research propose novel technique in web user data analysis for behavioural artifacts detection using machine learning (ML) architectures. Here web user access data has been collected and processed for noise removal and smoothening. Then this data feature is trained and selected for detection of malware activities using attribute ratio rule based auto encoder training. Then the selected data is classified using ensemble of spatio temporal Q-learning architectures. the experimental analysis is done for various dataset in terms of accuracy, F_measure, detection time, mean average precision, processing time, specificity. Proposed technique attained accuracy of 97%, F_measure of 86%, detection time of 75%, mean average precision of 63%, processing time of 77%, specificity of 85%.
We develop a machine learning-based pedestrian detection and alert system that can operate both during the day and at night using a visual camera, an infrared camera, and a radar sensor. The visible camera is used to ...
详细信息
We develop a machine learning-based pedestrian detection and alert system that can operate both during the day and at night using a visual camera, an infrared camera, and a radar sensor. The visible camera is used to detect pedestrians during the daytime while the infrared camera during the nighttime. Whereas the radar sensor is utilized to detect the presence of pedestrians including their range and directions of motion. We have developed and conducted actual experimentation of the system in a vehicle. We achieved an average accuracy of 98% based on our proposed multi-sensor data analysis using a deep learning algorithm that classifies a pedestrian’s presence during the day and at night and alerts the driver in a real-time monitoring system.
The design and implementation of efficient routing architectures is a critical aspect of modern communication systems. This paper proposes a modified VLSI-based router architecture that is optimized for high-speed dat...
The design and implementation of efficient routing architectures is a critical aspect of modern communication systems. This paper proposes a modified VLSI-based router architecture that is optimized for high-speed data transfer and low power consumption. The proposed architecture utilizes advanced routing algorithms and state-of-the-art VLSI design techniques to achieve a high level of performance and scalability. The performance of the design is evaluated through simulations. The simulation was carried out in a software called Xilinx and it is written using VHDL language. Design contains blocks called Arbiter, Cross bar and FIFO. The results show that the proposed architecture is able to achieve high throughput while maintaining a high level of scalability. This work is a significant step towards the development of high-performance communication systems.
This paper investigates the impact of data valuation metrics (variability and coefficient of variation) on the feature importance in classification models. data valuation is an emerging topic in the fields of data sci...
详细信息
This paper proposes a semi-supervised training approach for a direction-of-arrival (DoA) estimation based on a convolutional neural network (CNN). We apply a sparse recovery algorithm called optMGD-ℓ 1 -SVD on the tra...
详细信息
This paper proposes a semi-supervised training approach for a direction-of-arrival (DoA) estimation based on a convolutional neural network (CNN). We apply a sparse recovery algorithm called optMGD-ℓ 1 -SVD on the training dataset consisting of only unlabeled observed data to obtain binarized pseudo-spectra regarded as the CNN training targets (labels). The estimated DoAs are obtained at test time by performing peak picking on the CNN outputs. optMGD-ℓ 1 -SVD has been shown to perform well with a few sensors under low signal-to-noise ratio (SNR) conditions (up to −6 dB) by optimally reweighting the pseudo-spectra of ℓ 1 -SVD based on the application of group delay function on the pseudo-spectra of MUSIC. Since its hyperparameters are noise-sensitive, we assume that the SNR levels of the training dataset are known such that we can use the optimal ones. We also consider multi-condition training using data of multiple SNR levels to improve the robustness towards different noisy environments. We evaluated the trained networks, named optMGD-ℓ 1 -SVD-CNN and MGD-ℓ 1 -SVD-CNN, in terms of the average root-mean-square error and the resolution probability under low SNR conditions (up to −20 dB). We demonstrated that it performed well with a few sensors and snapshots, including at SNR levels unseen in the training data.
Since many years smartphones are utilised for human activity recognition (HAR), important healthcare recommendations and telemedicine. Deep learning (DL) and machine learning techniques are commonly employed in studie...
详细信息
The analysis of the sentiments can be determined if a piece of message or text or comment has a decent, negative or neutral feeling or sentiment. Analysis of sentiments is a subtype of natural language processing that...
详细信息
The analysis of the sentiments can be determined if a piece of message or text or comment has a decent, negative or neutral feeling or sentiment. Analysis of sentiments is a subtype of natural language processing that incorporates data extraction. For this purpose, numerous types of techniques and algorithms are used. In this paper used the previous twitter data for sentimental analysis and categorization. It is necessary for sentiment analysis to make use of and execute appropriate data visualization and preprocessing. For the analysis of positive or negative sentiments machine learning algorithm is used. In this work used the naive byes classifier for the analysis of the sentiments.
暂无评论