High-Performance Machine Learning

This course introduces Machine Learning (ML) from a system-oriented aspect. After a brief review of selected ML algorithms including both traditional and deep-learning-based, the course will provide an in-depth analysis of Google TensorFlow, including both architecture and components. Students will learn how ML algorithms can be efficiently implemented in TensorFlow in a scalable manner by utilizing TensorFlow’s infrastructure and understand various design tradeoffs between system abstractions, extensibility and performance. The next topic focuses on utilizing TensorFlow’s extensible architecture to develop custom operations on multi-core CPUs and GPUs that can be utilized interactively in Python though demonstrative examples.