جامعة المعرفة
المصدر المعتمد للمعلومات البحثية والمخرجات هو قاعدة بيانات سكوبس (Elsevier Scopus) مع تدقيق بيانات الباحثين من عمادة البحث العلمي.

Research outputs

المخرجات البحثية

مقالات ومراجعات وأوراق مؤتمرات وفصول كتب، مفهرسة برقم DOI ومربوطة بباحثي الجامعة.

2 مُخرَج نتائج البحث عن «gradient vanishing»

ورقة مؤتمروصول مغلق20244استشهاد· Semantic Scholar

Deep Learning-Based Intrusion Detection Model for Internet of Things Environment

IEEE International Conference on Innovations in High Speed Communication and Signal Processing (IHCSP) · 1-8

The proposed IoT intrusion detection model uses machine learning to compensate for data loss in convolutional neural networks' initial pooling layer and the gradient vanishing problem that recurrent neural networks face when processing long sequential input. Time series data for intrusion detection models is optimized using Gramme matrix field visualization. To reduce data loss, the SoftPool pooling layer replaced the original. The recurrent neural network uses a bidirectional gated recurrent unit with fewer parameters and a multi-head structure modeled after the self-attention mechanism. Baseline data will be TON-IoT data. The number of heads multi-head bidirectional gated recurrent unit, pooling layer topologies, and visualization approaches are experimental parameters. The simulation results show that a three-head structure, SoftPool for layer pooling, and for accuracy, precision, detection rate, F1 score, and false positive rate, multi-class classification works well with field and Gramme matrix visualization. The suggested model outperforms state-of-the-art models in accuracy, detection rate, F1 score, and false positive rate by 0.14, 0.09, 0.13, and 0.08. Modern machine learning improves IoT intrusion detection.

IoT intrusion detectionInternet of Things securitybidirectional gated recurrent unitmulti-head self-attention
مقالوصول مفتوح20234استشهاد· Semantic Scholar

Adaptive Rectified Linear Unit (Arelu) for Classification Problems to Solve Dying Problem in Deep Learning

International Journal of Advanced Computer Science and Applications · مجلد 14 (2)

A convolutional neural network (CNN) is a subset of machine learning as well as one of the different types of artificial neural networks that are used for different applications and data types. Activation functions (AFs) are used in this type of network to determine whether or not its neurons are activated. One non-linear AF named as Rectified Linear Units (ReLU) which involves a simple mathematical operations and it gives better performance. It avoids rectifying vanishing gradient problem that inherents older AFs like tanh and sigmoid. Additionally, it has less computational cost. Despite these advantages, it suffers from a problem called Dying problem. Several modifications have been appeared to address this problem, for example; Leaky ReLU (LReLU). The main concept of our algorithm is to improve the current LReLU activation functions in mitigating the dying problem on deep learning by using the readjustment of values (changing and decreasing value) of the loss function or cost function while number of epochs are increased. The model was trained on the MNIST dataset with 20 epochs and achieved lowest misclassification rate by 1.2%. While optimizing our proposed methods, we received comparatively better results in terms of simplicity, low computational cost, and with no hyperparameters.

leaky ReLUactivation functionsconvolutional neural networksMNIST classification