| An Innovative Evolutionary Computation Strategy for Optimizing Deep Learning Network |
| Paper ID : 1055-ICRSSSA-FULL (R1) |
| Authors |
|
shahera saad Ali Ali *1, Prof.Yehia Mostafa Helmy Helmy2, prof.Ibrahim Fathy Moawad Moawad3 1Al-Obour Higher Institute for Management and Informatics 2Head of Business Information Systems Department Faculty of Commerce Helwan University 3Ptofessor,Faculty of Computers and Information, Ain Shams University ,Egypt |
| Abstract |
| Deep learning is one of the subgroups of machine learning widely used in artificial intelligence (AI) fields such as remote sensing (RS) imagery and machine vision. RS is one of the most powerful techniques for understanding and recognizing the pattern of urban growth, and land use/land cover change (LULC) in a given area. Currently, most modern applications of pattern recognition are based on Machine Learning (ML) technologies. Deep learning convolutional neural network (CNN) is widely used as a big data analytics technique, particularly for clustering and/or classification of the RS imagery extracting high-level concepts from low-level features. However, Big data analytics problems are quite difficult to solve due to their large, high-dimensional, and dynamic properties. factually, most CNNs designs are still manually adjusted. Therefore, getting the best-performing CNN model is time-consuming and sometimes unattainable. Researchers recently started using Evolutionary Computation (EC) optimization algorithms to automatically adjust the hyperparameters of CNNs. EC-based optimization method plays a critical role in optimizing CNNs to reduce resource costs and accelerate performance while increasing accuracy. This article proposes a novel evolutionary algorithm named the Learner Performance-Based Behavior Algorithm (LPB) to optimize CNN automatically. The CNN-based-LPB model is proposed for recognizing the urban pattern to make effective and convenient urban pattern recognition. |
| Keywords |
| Keywords: Remote Sensing, Learner Performance-Based Behavior, Convolutional Neural Network, Optimization. Pattern Recognition. |
| Status: Accepted |
