Reproducible Data Science with Azure Machine Learning

Source: Reproducible Data Science with Machine Learning

YouTube Video

Overview

Being able to explain your own code a few months after you wrote it is hard. Imagine having to explain the decisions of some AI algorithm a few years after it run! However, it is relatively easy to set up your development workflow to make that possible, as long as you realize that the way we build ML and AI is fundamentally different from traditional software engineering.

In a nutshell, it is all about: Reproducible Research, Development and Deployment. It is made possible by a clever use of modern notebook environments, including Azure ML Compute Instances, as opposed to the more traditional IDEs, like Visual Studio Code.

Reproducible Workflow

1. Create a project to hold source, in-memory data, and history
2. Version Control with Git
3. Select a snapshot of package sources
4. Work in a developer notebook
5. Save and Export/Share your Findings
6. Push to Production

Resources

More Information:

Learn more about this way of using R with Rafal https://aka.ms/AIShow/UsingRwithRafal
Videos available at Tecflix https://aka.ms/AIShow/Tecflixvideos
Follow Rafal on LinkedIn https://aka.ms/AIShow/RafalonLinkedIN
Follow Rafal https://twitter.com/rafaldotnet

Backlinks:

list from [[Reproducible Data Science with Azure Machine Learning]] AND -"Changelog"