Knowledge Graph — Coursera Notes › Organizations › Microsoft › Azure › Azure ML Service

Azure Machine Learning workspace

service · part of Azure ML Service

The Azure Machine Learning workspace is the central hub for managing all machine learning activities, including experiments, datasets, models, and compute resources. It provides a unified interface for organizing, securing, and scaling AI/ML projects. Key components include compute resources (compute instances, compute clusters, inference clusters), datastores, datasets, experiments, and model registration. Benefits include centralized management, scalability, collaboration via role-based access control, and experiment tracking with model versioning.

• Compute instances: development environments for Jupyter notebooks, testing, and small datasets; ideal for interactive work.
• Compute clusters: scalable VM clusters for large-scale training jobs; distribute training across multiple nodes.
• Inference clusters: optimized for deploying models to serve real-time or batch predictions.

Datastores provide secure connections to Azure storage services (e.g., Azure Blob Storage, Azure Data Lake) for managing raw data. Datasets are structured views of data within a datastore; they can be registered for reuse across experiments. Azure supports tabular datasets (CSV, SQL tables) and file datasets (images, text files).

The workspace also supports automated machine learning (AutoML) for automatically training and tuning models, and pipelines for orchestrating reusable workflows. Additionally, it integrates with Azure DevOps and GitHub for CI/CD, and provides a designer for drag-and-drop model building.

Inside Azure Machine Learning workspace (3)

• Azure ML Data Import — Jobs to bring data from external sources into Azure ML can run once or on a schedule.
• Azure ML Workspace Creation — To create an Azure ML workspace, use the Azure portal: log in, select 'Create a resource', search for 'Machine learning', click on the Azure Machine Learning service, and click 'Create'.
• Data Drift Detection — Data drift detection: A preview feature (now deprecated) that monitors changes in data distribution between training and inference data.

Also known as: Azure Machine Learning workspace, Machine Learning workspace

This is the text view of an interactive 3D knowledge graph — open this page with JavaScript enabled to explore it visually.