Deep learning algorithm for data classification with hyperparameter optimization method

Abstract

The availability of data is an important thing in the decision-making process. With the trend of using Machine Learning and Data Mining, humans are facilitated in learning and finding hidden knowledge in the data. One technique used in the classification of data is the Deep Learning method. Deep Learning is widely used because of its ability to solve problems with satisfying results. In other words, Deep Learning has shown to be the one of the most powerful and widely used methods. However, the key of success depends on the setting the right parameters. With this reason, we need a way to optimize the use of hyperparameter in Deep Learning. The method commonly used in this case is Grid Search. However, this method requires a long time and high computation. While other methods such as Random Search and Bayesian Optimization, both are very reliable and efficient compared to Grid Search. The results of the experiments conducted in this study indicate that Bayesian Optimization is better than Random Search in performing hyperparameter optimization processes in Deep Learning.