Deep Neural Networks (DNNs) are powerful models that can solve different tasks as speech, images,
and natural language understanding. In recent years, they have outperformed classical Machine Learn-
ing (ML) algorithms on numerous tasks. The problem with deep learning algorithms is designing them
manually and tuning their parameters, which consumes time and requires human effort. To solve this is-
sue, researchers have proposed Neural Architecture Search (NAS), which automatically designs a DNN
for a specific task. NAS automates network architecture engineering. It aims to find a network topology
to achieve the best performance on a specific task.

This study proposes a NAS approach based on evolutionary algorithm for searching Convolutional
Neural Networks (CNNs) architectures. For solution representation, a continuous encoding scheme is
also proposed.