Manta Ray Foraging Optimization for Hyper-Parameter Selection in Convolutional Neural Network
Omar Adil Kamil, Shaymaa W. Al-Shammari · IOP Conference Series Materials Science and Engineering · 2020
Abstract Convolutional neural networks (CNNs) have been attracting attention as one of the most common deep learning techniques used for different applications like images classifications, objects recognition, face recognition, etc. The performance and efficiency of the CNN model depend directly on their hyper-parameter, which must be selected by an expert or using one of the models that proposed and improved. This makes it very important to determine the optimal hyper-parameters. In this work, the Manta Ray Foraging Optimization (MRFO) algorithm is used to select CNN’s Hyper-Parameter. We demonstrate that MRFO efficiently explores the solution space. This allowing CNN with simple architecture to achieve a good classification accuracy over Cifar_10 dataset. In the presented experiment Cifar_10 dataset used as the benchmark data sets. By optimizing CNN hyper-parameters with MRFO algorithm and comparing the obtained results with other CNN, it was approved that the accuracy was improved.