Convolutional neural networks are able to perform a hierarchical learning process starting with local features. However, a limited attention is paid to enhancing such elementary level features like edges. We propose and evaluate two wavelet-based edge feature enhancement methods to preprocess the input images to convolutional neural networks. The first method develops feature enhanced representations by decomposing the input images using wavelet transform and limited reconstructing subsequently. The second method develops such feature enhanced inputs to the network using local modulus maxima of wavelet coefficients. For each method, we have developed a new preprocessing layer by implementing each purposed method and have appended to the network architecture. Our empirical evaluations demonstrate that the proposed methods are outperforming the baselines and previously published work with significant accuracy gains.
Single object tracking is a well-known and challenging research topic in computer vision. Over the last two decades, numerous researchers have proposed various algorithms to solve this problem and achieved promising results. Recently, Transformer-based tracking approaches have ushered in a new era in single object tracking due to their superior tracking robustness. Although several survey studies have been conducted to analyze the performance of trackers, there is a need for another survey study after the introduction of Transformers in single object tracking. In this survey, we aim to analyze the literature and performances of Transformer tracking approaches. Therefore, we conduct an in-depth literature analysis of Transformer tracking approaches and evaluate their tracking robustness and computational efficiency on challenging benchmark datasets. In addition, we have measured their performances on different tracking scenarios to find their strength and weaknesses. Our survey provides insights into the underlying principles of Transformer tracking approaches, the challenges they face, and their future directions.
Phishing, a well-known cyber-attack practice has gained significant research attention in the cyber-security domain for the last two decades due to its dynamic attacking strategies. Although different solutions have been exercised against phishing, phishing attacks have dramatically increased in the past few years. Recent studies have shown that machine learning has become prominent in the present antiphishing context, and the techniques like deep learning have extensively improved anti-phishing tools' detection ability. This paper proposes PhishDet, a new way of detecting phishing websites through Longterm Recurrent Convolutional Network and Graph Convolutional Network using URL and HTML features. PhishDet is the first of its kind, which uses the powerful analysis and processing capabilities of Graph Neural Network in the anti-phishing domain and recorded 96.42% detection accuracy, with a 0.036 false-negative rate. It is effective against zero-day attacks, and the average detection time which is 1.8 seconds could also be considered realistic. The feature selection of PhishDet is automatic and occurs inside the system, as PhishDet gradually learns URLs and HTML content features to handle constantly changing phishing attacks. This has outperformed similar solutions by achieving a 99.53% f1-score with a public benchmark dataset. However, PhishDet requires periodic retraining to maintain its performance over time. If such retraining could be facilitated, PhishDet could fight against phishers for a more extended period to safeguard Internet users from this Internet threat.
Unconstrained growth of synaptic connectivity and the lack of references to synaptic depression in Hebb's postulate has diminished its value as a learning algorithm. While spike timing dependent plasticity and other synaptic scaling mechanisms have been studying the possibility of regulating synaptic activity on neuronal level, we studied the possibility of regulating the synaptic activity of Hebb's neurons on dynamic stochastic computational synapses. The study was conducted on fully connected network with four artificial neurons where each neuron consisted of thousands of artificial stochastic synapses that are modeled with transmitters and receptors. The synapses updated their stochastic states dynamically according to the spike arrival time to that synapses. The activity of these synapses was regulated by a new stability promoting mechanism. Results support the following findings: (i) the synchronous activity between presynaptic (cell A) and postsynaptic (cell B) neuron increases the activity of A. (ii) Asynchronous activation of these two neurons decreases A's activity if one of the following conditions are satisfied (a). if activity of the other presynaptic neurons of the postsynaptic neuron B is asynchronous with the A's activity or (b) if B is in a depressed state when activity of presynaptic neuron A isincreased. (iii) the introduced stability promoting mechanism exhibited similar to the Homeostatic synaptic plasticity process and encouraged the emergence of Hebb's postulate and its anti-Hebbian mechanisms. Further, we demonstrated the metabolic changes that could occur inside Hebb's neurons when such an activity takes place on a dynamic stochastic neural network.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.