Abstract
Analysis of large gene expression datasets for cancer classification is a crucial task in bioinformatics and a very challenging one as well. In this paper, we explore the potential of using advanced models in machine learning namely those based on deep learning to handle such task. For this purpose we propose a deep feed forward neural network architecture. In addition, we also investigate other classical yet very popular machine learning classifiers namely, support vector machine, naive bayes, k-nearest neighbours and shallow neural networks. The main objective is to appreciate the extent to which they are able to deal with the increasing size of these datasets. We conducted our experimental study using a high-performance computing platform with 32 compute nodes, each consisting of two Intel (R) Xeon (R) CPU E5-2650 2.00 GHz processors. Each processor is made up of 8 cores. Five data sets available at the omnibus library have been used to test the five models . Experimental results show the effectiveness of deep learning and its ability to deal with large scale data.