Azure Machine Learning: the Overview

In this video, we will provide an introduction to Azure Machine Learning, including how to provision the service and a high-level overview of the primary concepts involved.
LINKS AND INFO
Catallaxy Services - www.catallaxyservices.com
Azure Machine Learning resources and assets -- learn.microsoft.com/azure/mac...
Chicago parking tickets dataset -- cspolybasepublic.blob.core.wi...
Chicago parking tickets database -- sqlsunday.com/2022/12/05/new-...

[@rickpack6700]

Studying for the Azure ML certification and appreciate this video, Kevin!

[@maguedoudasamiabelhaddad2183]

Amazing video dude! Thank You.

[@uliO]

Thanks - great overview.

[@rodrigocarballo4478]

This s great overview ... Thank you

[@haemonyang6278]

Great video.

[@tarsala1995]

Thanks for sharing the content. I need those bulky clusters for my PhD experiments.

[@srimanthn7752]

Great overview

[@palgun.]

Thanks

[@misael6506]

Great video! I wonder if Azure ML is charged by the processing time and VM configuration.

[@KevinFeasel]

It is. The pricing details focus on VM compute because that's the largest percentage by far for most customers: azure.microsoft.com/en-us/pricing/details/machine-learning/
The longer form of what you're paying for:
- Data stored in Azure Blob Storage and disk for the compute instances ($0.02 to $0.23 USD per GB)
- VM compute for your compute instances and compute clusters *when they are running*. Stopped VMs don't incur charges, and this means more efficient code finishing jobs faster on smaller VMs leads to a smaller bill.
- VM compute costs if you deploy the model using Azure Container Instances or Azure Kubernetes Service
- Storage of images in Azure Container Registry. This is usually quite inexpensive. You can often get away with the Basic tier, which is $0.17 USD per day. Standard tier is $0.67 per day.
- There are other fractional charges for Azure Key Vault, Application Insights, and Log Analytics. I don't even bother with Key Vault pricing as it's so little. App Insights and Log Analytics can get up there in price, but not for AML cases--the service just doesn't write enough to them to rack up an appreciable bill.
- You might also incur networking fees if you are consistently sending large amounts of data into and out of Azure regions. I almost never think about this because it's typically such a low number, especially for ML scenarios.
Aside from the core service costs, there's no additional fee to use Azure ML resources, e.g., some "number of trained models" or "number of API calls" fee. So the rule of thumb is, you're paying for VMs when they're running and data in your storage accounts, and that's going to be ~99% of your AML bill.

[@misael6506]

Thank you very much for your great answer! I appreciated it!@@KevinFeasel

[@mehdikhezrian2257]

Could you please not use the background music? it's very distracting :(