Optimize and Manage Models
Lecture
John Ellis
Intermediate
1 h 23 m
2020-09-18
Lecture Overview
What if Azure Machine Learning could do the hard work of model selection and optimization for you? What if it also could automatically explain the “why’s” of the model’s prediction process? Well it can! In this course we explore a few powerful features of Azure Machine Learning. In part one we will learn about Automated Machine Learning which makes algorithm selection a breeze. In part two we explore Hyperdrive, a feature that helps us find optimal Hyperparameter values. Then this course will introduce you to Interpretability in Azure Machine Learning. Finally, we learn about a powerful new feature in the platform called Data Drift Detection. This course covers one of the objective areas that help you in preparing for the DP-100: Designing and Implementing a Data Science Solution on Azure exam.

Related Learning Path(s):
DP-100: Designing and Implementing an Azure Data Science Solution on Azure
Objectives
  • Describe and understand Azure Automated Machine Learning (AutoML)
  • Describe and understand Hyperdrive and the problem of the hyperparameter search space
  • Understand model interpretability in Azure Machine Learning
  • Understand and detect drift on data sets
Lecture Modules
This module covers some of the most exciting features within Azure Machine Learning and is divided into 4 parts and one demonstration. We will begin by exploring the concept of Automated Machine Learning and take a look at what tooling Microsoft has created for us to make use of this new and exciting innovation in machine learning. We then discuss featurization which is the heart of successful machine learning. We then move on to review the types of compute targets available in Azure and discuss how the choice of compute target affects the product features available to us from automated machine learning. Finally we will learn about the concept of a Primary Metric and discuss how this one piece of logged information is what makes Auto ML possible.
This module is divided into two parts followed by a demonstration of what we’ve learned. We will first learn exactly what a hyperparameter is and why choosing them can be so problematic. Then we will introduce and learn about Azure Machine Learning’s answer to this problem: Hyperdrive. Finally we will pull together our new understanding and demonstrate the construction, execution and analysis of a Hyperdrive Experiment in Azure Machine Learning.
This module is divided into two parts followed by a demonstration of what we’ve learned. First, we will explore the concept of model interpretability and why it is important. Next, we will discuss what is meant by feature importance. Finally we will demonstrate model interpretation in practice by using the Python SDK to explore why one of our previous models made the choices it did.
This module is divided into three parts with one demonstration. We will explore how to register trained models with the workspace and how they are versioned.
We then move on to understand how we can monitor our models through data collection. Third, we explore another exciting feature of Azure Machine Learning called Data Drift Detection. Finally we demonstrate these technologies by implementing data drift monitoring on a model and it’s output via the Python SDK.
Try Risk Free

Start a free trial

Skill Me Up subscriptions include unlimited access to on-demand courses with live lab lab environments with our Real Time Labs feature for hands-on lab access.

Subscription Benefits
  • Access to Real Time Lab environments and lab guides
  • Course Completion Certificates when you pass assessments
  • MUCH MORE!