Deep Learning and parallelization of Meta-heuristic Methods for IoT Cloud

Khelili, Mohamed Akram (2022) Deep Learning and parallelization of Meta-heuristic Methods for IoT Cloud. Doctoral thesis, Université de mohamed kheider biskra.

Text Khelili_Mohamed_Akram.pdf Download (3MB)

Abstract

Healthcare 4.0 is one of the Fourth Industrial Revolution’s outcomes that make a big revolution in the medical field. Healthcare 4.0 came with more facilities advantages that improved the average life expectancy and reduced population mortality. This paradigm depends on intelligent medical devices (wearable devices, sensors), which are supposed to generate a massive amount of data that need to be analyzed and treated with appropriate data-driven algorithms powered by Artificial Intelligence such as machine learning and deep learning (DL). However, one of the most significant limits of DL techniques is the long time required for the training process. Meanwhile, the realtime application of DL techniques, especially in sensitive domains such as healthcare, is still an open question that needs to be treated. On the other hand, meta-heuristic achieved good results in optimizing machine learning models. The Internet of Things (IoT) integrates billions of smart devices that can communicate with one another with minimal human intervention. IoT technologies are crucial in enhancing several real-life smart applications that can improve life quality. Cloud Computing has emerged as a key enabler for IoT applications because it provides scalable and on-demand, anytime, anywhere access to the computing resources. In this thesis, we are interested in improving the efficacity and performance of Computer-aided diagnosis systems in the medical field by decreasing the complexity of the model and increasing the quality of data. To accomplish this, three contributions have been proposed. First, we proposed a computer aid diagnosis system for neonatal seizures detection using metaheuristics and convolutional neural network (CNN) model to enhance the system’s performance by optimizing the CNN model. Secondly, we focused our interest on the covid-19 pandemic and proposed a computer-aided diagnosis system for its detection. In this contribution, we investigate Marine Predator Algorithm to optimize the configuration of the CNN model that will improve the system’s performance. In the third contribution, we aimed to improve the performance of the computer aid diagnosis system for covid-19. This contribution aims to discover the power of optimizing the data using different AI methods such as Principal Component Analysis (PCA), Discrete wavelet transform (DWT), and Teager Kaiser Energy Operator (TKEO). The proposed methods and the obtained results were validated with comparative studies using benchmark and public medical data.

Actions (login required)

View Item