Open Standards for Machine Learning Deployment
summary
Compares practical deployment formats: PMML, PFA and ONNX, and how they map to scikit‑learn, Spark ML, TensorFlow, Keras and PyTorch. Reports strengths and limits — PMML excels for classical models but lacks generic tensors and deep‑learning support, PFA is flexible for scoring but has limited ecosystem and GPU support, and ONNX targets deep‑learning inference across runtimes but is mainly static and not for ongoing training. Recommends using a portfolio of standards and notes tooling and licensing gaps that block seamless cross‑team deployment.
AI-written summary · about
chapters (13)
- 0:00Start
- 2:59What is Machine Learning?
- 3:51Intelligent Systems
- 5:12Perception
- 5:46In reality the workflow spans teams...
- 9:54What, Where, How?
- 10:56What is a "model"?
- 11:51Pipelines, not Models
- 12:09Challenges
- 15:27Containers are The Solution... right?
- 17:16Why a standard?
- 18:37Predictive Model Markup Language
- 20:06PMML-Format overview
related talks
33mOpen Standards for Machine Learning Deployment
41mOpen standards for machine learning deployment
34mOpen Standards For Machine Learning Model Deployment
48mDeploying Machine Learning Models in Practice
52mMachine Learning Platforms
54mOne Universal Language for All Machine Learning Models
46mProduction Model Deployment
45mMaking Machine Learning Applications Fast and Simple with...