A Hyper-parameter Optimization for Deep Neural Network using an Island-based Genetic Algorithm
Rashrita rattanavorragant, Yutana Jewajinda · 2019
We present a method using an island-based genetic algorithm to automatically optimize the hyper-parameters of deep neural networks (DNNs). The method consists of two steps: hyper-parameter search and complete DNNs training. To demonstrate our method, we use deep convolutional neural networks (DCNNs). We encode structure and hyper-parameters of a DCNN as a chromosome and then apply an island-based genetic algorithm to optimize the structure and hyper-parameters. To evaluate the proposed method, we implemented an image classification task using CIFAR-10 dataset. The experimental results show that the proposed method can efficiently search for the hyper-parameters of DCNNs.