This roadshow was organized by NVIDIA. There were a few NVIDIA experts invited from the USA to give an overview about Deep Learning, the applications of it and how we can utilize NVIDIA GPU and CUDA framework to speedup Deep Learning. This is a very interesting roadshow because it is very hands-on. There are two hands on session during the roadshow, teaching us the basics of Deep Learning. The hands on session are introductions to Recurrent Neural Networks (RNN) and Classifying Whale Heads using DIGITS. DIGITS is a web-based deep learning tool that is very useful and flexible. It is also simple to use.
The roadshow starts with an introduction to Deep Learning and how it is gaining popularity nowadays. Deep Learning is basically a large Neural Networks such as Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN). Most of the applications that they mentioned revolved around self-driving car, Project DAVE. DARPA Autonomous Vehicle (DAVE) is a self driving car aiming to create a robust system for driving on public roads. The car was trained using NVIDIA DevBox and Torch 7 (Machine Learning Library). The self-driving cars was also shown to be able to identify road signs, landmarks and was able to drive in bad weather conditions (rain, snow). Below is a short video showing the demo of NVIDIA’s self driving car. More information on Project DAVE can be found in their blog here.
Later, they continued to talk about the SDK framework that had been developed by NVIDIA such as CUDA and the supercomputers that are efficient for Deep Learning. After lunch, we have Professor Tom Drummond from the Department of ECSE, Monash to give a talk about his research and how he apply Deep Learning using GPU computing for robotics and computer vision applications, especially the Bionic Eye Project. He also stressed the importance and usefulness of Deep Learning. After the talk, we have the hands on session, starting with the DIGITS lab, followed by the introduction to RNN.
After this roadshow, I gain more interest and understanding in Deep Learning. Deep Learning is definitely a useful method for a lot of applications, including time series. In the future, I would like to apply Deep Learning techniques to my research. I believe that it will be very useful for my projects.
Time: 8.30am – 4.30pm