Abstract:

Deep Learning is probably the area of Machine Learning which currently arouses the greatest interest in both academic and research spheres. However, its introduction in Computer Science and Computer Engineering curricula is often impaired because the background needed to understand its basis would required too many teaching hours.

In this paper we proposed an approach for teaching Deep Learning which avoids the need of a theoretical basis in the field. It is based on the NVIDIA framework DIGITS (Deep Learning GPU Training System). DIGITS is accessed via a web browser and offers a user-friendly interface. Moreover, it provides all the necessary tools to facilitate the introduction of the students in the field.

As we show in the paper, using a graphical environment to introduce the subject allows a fast learning on the key concepts. In addition, following a practical approach make the learning more appealing for students.

The end result is that students get introduced into Deep Learning without devoting many class hours. This allows students being able to decide whether the subject actually interest them or not, serving also as an initial basis for future training.

Keywords:

Deep Learning, Hands-on training, Computer Science and Engineering.