The Cherenkov Telescope Array (CTA) will be the largest ground-based, gamma-ray observatory. CTA will detect the signature of gamma rays and cosmic rays hadrons and electrons interacting with the earth atmosphere. Making the best possible use of this facility requires to be able to separate events generated by gamma rays from the particle-induced background. Deep neural networks produced encouraging results, but so far there has been no evaluation of their performance for gamma/hadron separation with respect to well established approaches. In this paper we compare convolutional neural networks and a standard analysis technique, namely boosted decision trees. We compare the performance of the two techniques as applied to simulated observation data. We then looked at the Receiver Operating Characteristics (ROC) curves produced by the two approaches and discuss the similarities and differences between both. We find that neural networks outperformed classical techniques under specific conditions.