Responsible AI with Azure Machine Learning
Tools and methods to understand, protect, and control your models
In this paper, we’ll share some of the tools and methods that
help our IT teams honor our AI principles. The same tools we use internally are available to everyone in Azure Machine Learning, which provides state-of-the-art capabilities to help users understand, protect, and control their data, models, and processes.
Three Responsible Machine Learning pillars
In Azure Machine Learning (Azure ML), we focus on three categories of responsible ML capabilities, which ladder up to our company-wide AI principles. These categories are:
Understand: Gain visibility into your models, explain model behavior, and detect and mitigate model bias.
Protect: Apply differential privacy techniques to protect sensitive data and prevent leaks. Encrypt data and build models in a secure environment to maintain confidentiality.
Control: Use built-in lineage and audit trail capabilities and document model metadata to meet regulatory requirements.