Looking Under the Hood of Deep Neural Networks

Abstract

Treating deep neural networks as black boxes and using them as-is from a toolbox could potentially lead to sub-optimal performance. Increasingly machine learning researchers have to be more aware of the computational workloads entailed by their models and how to optimize for them. In this talk, I will describe three different pieces of our recent work with deep convolutional networks and their variants in improving inference performance across a variety of tasks like object detection, identification, tracking, etc. These studies demonstrate the need for peeling back the cover and paying attention to the computation even when using standard models. Biography Balaram Ravindran is a professor at the Department of Computer Science and Engineering, and the head of the Robert Bosch Centre for Data Science and AI at the Indian Institute of Technology Madras. His current research interests span the broader area of machine learning, ranging from Spatio-temporal Abstractions in Reinforcement Learning to social network analysis and Data/Text Mining. Much of the work in his group is directed toward understanding interactions and learning from them. 1 2018 IEEE 25th International Conference on High Performance Computing (HiPC) DOI 10.1109/HiPC.2018.00009