A Novel Approach for Hyperspectral Image Classification using Bat Algorithm to Optimize a CNN Classifier

We propose a novel hybrid classifier for hyperspectral images using Bat Algorithm (BA) to optimize the architecture of a Convolutional Neural Network (CNN). BA is applied by minimization of the CNN cross-entropy on the validation set. The objective function is represented as a matrix containing the parameters which define the CNN architecture. The proposed BAT-CNN classifier is evaluated on three hyperspectral datasets: Indian Pines, Pavia University and Salinas. The experiments lead to a better accuracy for the proposed hybrid classifier by comparison to the standalone CNN classifier, for each of the three considered datasets.