Skip to content

Latest commit

 

History

History
36 lines (35 loc) · 2.68 KB

2.3. SaaS services in Azure - Azure Machine Learning.md

File metadata and controls

36 lines (35 loc) · 2.68 KB
Raw

Azure Machine learning

  • Azure Machine learning is an end-to-end data science and analytics solution that's integrated into Azure.
  • Built on top of open source technologies: Jupyter Notebook, Conda, Python, Docker, Apache Spark, and Kubernetes (also from Microsoft, e.g. Cognitive Toolkit)
  • It allows users to develop experiments as well as deploy data and models via the cloud.
  • Its composed of
    • Azure Machine Learning Workbench
      • Desktop application that includes command-line tools.
      • It allows users to help manage learning solutions via data ingestion and preparation, model development, experiment management,
    • Azure Machine Learning Experimentation Service
      • Helps handling the implementation of machine learning experiments
      • Provides project management, roaming, sharing, and git integration to support the Workbench.
      • Allows implementation of services across a range of environment options such as Local native, Local Docker container, or Scale out Spark cluster in Azure.
      • Creates Virtual environments for scripts to provide an isolated space with reproducible results.
        • Documents run history information
        • Visually displays the information so you can select the best model from your experiments.
    • Azure Machine Learning Model Management Service
      • Provides users the ability to deploy predictive models into a range of environments.
      • Information on models, such as the version and lineage, is notated from training runs throughout the deployment.
      • The models themselves are registered, managed, and stored in the cloud.
    • MMLSpark (Microsoft Machine Learning Library for Apache Spark)
      • Open-source Spark Package providing data science and Deep Learning tools for Apache Spark.
      • MMLSpark allows users to create robust, analytical, and highly scalable predictive models for large image and text datasets.
    • Visual Studio Code Tools for AI
      • Extension used with Visual Studio code that allows you to test, build, and deploy AI and Deep Learning solutions.
      • It contains various integration points from Azure Machine learning.
      • E.g. visualization of run history that displays the performance of training runs, select targets for your scripts to execute.
  • Fully support various open source technologies, such as scikit-learn, TensorFlow, and more.
  • Traditional BI flow: (value & amount of information increases in each step)
    • Descriptive analytics: What happened?
      • Leads to hindsight
    • Diagnostic analytics: Why did it happen?
      • Leads to insight
    • Predictive analytics: What will happen?
      • Leads to optimization & foresight
    • Prescriptive analytics: How can we make it happen?