Automated ML (Machine Learning) AMA (Ask Microsoft Anything)

Hi Leslie, This is a link for what's new for Azure ML (v2): https://docs.microsoft.com/en-us/azure/machine-learning/concept-v2

This is what's new for AutoML:

- AutoML in Azure ML CLI v2 Preview: Until now, AutoML didn’t support any CLI (AutoML was not part of Azure ML CLI v1). AutoML in Azure ML now offers CLI v2 which allows users to define AutoML standalone jobs using a YAML file and use CLI command create to create the AutoML job in Azure ML compute. The use of command-line makes deployment and automation of workflows easier since they can be invoked from any offering/platform which allows to call the command line. This is an ideal scenario for MLOps and rapid testing.

- AutoML in Azure ML Python SDK v2 Preview: AutoML in Azure ML now offers updated functions and classes in Python SDK v2 package. This new version provides a leaner and more structured Python functions with shorter and leaner signatures adjusted per AutoML task. This provides significant improvements in Python coding usability compared to SDK v1.

- AutoML Standalone Jobs Public Preview and AutoML in Pipeline Jobs Private Preview: In this public preview, AutoML standalone jobs are supported. AutoML in Pipelines is currently supported in Private Preview and will be released publicly soon.

- New AutoML features in v2: In addition to supporting the new CLI v2 and SDK v2, AutoML is also releasing new machine learning tasks plus additional Forecasting improvements, mentioned below:

Supported AutoML tasks in v2:

AutoML for Tabular Data

- Classification task

- Regression task

- Time Series Forecasting task

AutoML for Images (Computer Vision)

- Image Classification multiclass (**New**)

- Image Classification multilabel (**New**)

- Image Object Detection (**New**)

- Image Segmentation (**New**)

AutoML NLP (Natural Language Processing)

- NLP Text Classification multiclass (**New**)

- NLP Text Classification multilabel (**New**)

- NLP Text NER (Named Entity Recognition) (**New**)

Improvements in AutoML Time Series Forecasting:

- Forecasting horizon visualization (**New**)

- Frequency data aggregation (**New**)

- Automatic Time Series ID detection (**New**)

- Back testing script (**New**)

- Custom cross validation step size (**New**)

AutoML model’s Code Generation: ‘AutoML Code Generation’ makes AutoML a ‘White Box’ AutoML solution by allowing the user to select any AutoML trained model (best or child model) and generate the Python training code that created that specific model. Then, explore, customize, and retrain the model using Python before deploying to Azure ML Endpoints or your selected inference execution environment.