Azure Machine Learning is an online platform for developing, testing, and deploying machine learning models in the cloud at scale.
Built on the Microsoft Azure cloud platform, Azure Machine Learning allows you to:
Set up scalable virtual machine clusters for training AI workloads.
Store data and manage services to read data from a wide range input sources.
Visually design advanced machine learning models.
Register new models and the data on which they were trained.
Monitor the performance of training and production workloads.
Publish fully-trained models for real-time and batch inferencing.
We will cover all these features in detail in the upcoming course assignments.
Azure Machine Learning Workspaces
A workspace is a container for a set of machine learning experiments, the corresponding data files, a list of compute targets to train and run models on, and various other assets associated with a machine learning project.
Workspaces group a set of related machine learning resources. You can use workspaces to manage machine learning assets based on projects, deployment environments, teams, or anything else that works for you.
The assets in a machine learning workspace include:
Compute targets that identify virtual machine clusters for development, training, and deployment.
Datasets for experimentation and model training.
Python notebooks with shared code and documentation.
Machine learning experiments with run history, metrics, and outputs.
Pipelines that define orchestrated multi-step machine learning processes.
Machine learning models that you have trained.
When deploying your own Azure Machine Learning workspace, there are two things you need to consider in advance.
First of all, you need to decide where to deploy your workspace. It will be created in a specific Azure region which specifies the location of the data center in which the workspace resources will be hosted. You should ensure that the region where you create the workspace will support the other resources you will need.
For example, you might want to train your machine learning models on virtual machines with heavy-duty GPUs (the NC-series virtual machines in Azure). You should ensure that the region where your workspace is hosted also supports these virtual machines.
Second, the Azure Machine Learning workspace is available in two editions:
Enterprise edition: contains all features.
Basic edition: includes core features but does not contain the visual designer nor the graphical tools for automated machine learning and data drift monitoring.
For this course you will use the visual designer and therefore you'll need to install the Enterprise edition.
Azure Machine Learning studio
You can manage the assets in your Azure Machine Learning workspace with a dedicated interface called the Azure Machine Learning Studio.
The Azure Machine Learning studio is a web-based application for managing an Azure Machine Learning workspace. The studio allows you to create, manage, and view all of the resources in your workspace and provides the following tools:
The model designer: A drag and drop interface for no-code machine learning model development.
The Auto ML service: A tool that enables you to automatically train a model by discovering the best combination of algorithms and data preprocessing techniques.
We will be using these tools in the upcoming assignments in this course to build all kinds of cool machine learning systems.
Azure ML Studio Code Collection
This course will introduce you to Azure Machine Learning Studio and get you up to speed with building and training machine learning models at cloud scale.