Training Deep Neural Networks with Tensor Processing Units

Teaching machines to learn from data has made tremendous improvements with newer hardware and algorithms that better use these capabilities. Faster hardware enables running more experiments in a shorter period of time, providing more learning opportunities and evaluating newer network architectures in a shorter period of time. In this talk, we will explore how Google’s TPUs can be used to train deep neural network architectures in a limited amount of time. We will also have a hands-on session to understand how to use TPUs for a training task.

Karthik Muthuswamy

Senior Data Scientist | SAP

Working on machine learning research to develop intelligent applications for enterprise software.

Experience with leveraging machine learning for developing a wide range of applications. Conducts enablement sessions for software developers to jumpstart their journey on developing applications that use machine learning. Actively learning about human-centered machine learning and eager to bring it into mainstream machine learning application development.


Mar 16 2019


04:00 PM - 05:00 PM