Open Standards for Machine Learning Deployment

Explore the critical yet often overlooked aspect of deploying machine learning models to production systems in this informative conference talk. Delve into the current state of ML deployment using open-source, standardized formats, focusing on the "last mile" of the process. Learn about various available options, including PMML, PFA, and ONNX, and understand how these integrate with popular ML libraries such as scikit-learn, Spark ML, TensorFlow, Keras, and PyTorch. Gain insights into the challenges of ML deployment, the role of containers as a potential solution, and the importance of open standards in the field. Discover the intricacies of Predictive Model Markup Language, Portable Format for Analytics, and Open Neural Network Exchange (ONNX), along with the tools that support ONNX.

Intro
What is Machine Learning?
Intelligent Systems
In reality the workflow spans teams...
What is a "model"?
Challenges
Containers are The Solution... right?
Why a standard?
Predictive Model Markup Language
Portable Format for Analytics
Open Neural Network Exchange ONNX
ONNX Supported Tools