You should be using Azure Machine Learning designer
A modern update with a familiar look and feel gives data scientists the best of both worlds
I have been a strong advocate of Microsoft’s Azure Machine Learning Studio (classic) since its release in 2015. It is a great tool, parts of it remain relevant and it is an excellent platform for students new to machine learning.
It has an outstanding and its intuitive visual interface make it easy to use. It can be extended to be as powerful as you need it to be with R and Python scripts. Its ties to R are strong — this reflects Microsoft’s investment in R at the time. Microsoft published a slew of galleries and examples that made Azure ML Studio an exceptional machine learning resource. It was — and still is — a go-to place to see how someone else has solved a problem in a way that is easy to understand.
Fast forward to late 2020 and the world has moved on. Microsoft has launched an impressive range of machine learning services under the Azure banner, with strong ties to Python libraries. They have now made them essentially free. The focus of machine learning has tilted slightly towards Python and away from R. Microsoft has started to deprecate its development in Azure Machine Learning Studio (classic) in favour of its new suite of machine learning services. The direction for R is through integration with SQL Server (named SQL Server Machine Learning Services).
Is it me, or are the names confusing?
- Azure Machine Learning Studio (classic) — A visual machine learning environment
- Azure Machine Learning Services — Azure-based API and integrated Python libraries for the full machine learning lifecycle
- SQL Server Machine Learning Services — R and Python machine learning in SQL Server
Come on Microsoft, bring back names we can relate to. Like Clippy.
With all of this, faithful users were left with a choice: do I want ease of use or tight integration with Azure? You could have flexibility in Azure environments but were limited to code and the learning curve that comes with that. Or you could have easy development with a brilliant visual interface, but were limited in the Azure environments in terms of scale, compute, regions and storage for deploying models.
Enter the new visual interface for Azure Machine Learning known as (I think) Azure Machine Learning designer which is part of the new and equally-confusingly-named Azure Machine Learning Studio. If you are familiar with Azure Machine Learning Studio (classic), you will know how to use the new interface. It looks exactly the same and it shares many features with its predecessor.
Look at the screen grab below of the new Azure Machine Learning designer interface: the interface we love is back — with upgrades!
Things are not quite the same. The ties with R have been replaced by stronger — and better — ties to Python. You have full control over your compute environment, and you have access to the full Azure run history of experiments. So you have flexibility with Azure, plus a great visual coding interface which you probably already know.
Conclusion
To sum up, the new visual interface for Azure Machine Learning Service is a great addition to Microsoft’s Azure machine learning suite. You get the best of both worlds: power, flexibility and an intuitive interface plus all the bells and whistles from the Azure Machine Learning suite. These include automated machine learning, interpretable machine learning and model monitoring, all important capabilities for a modern machine learning suite.
