Statistical Machine Learning with Microsoft ML

MicrosoftML is an R package for machine learning that works in tandem with the RevoScaleR package. (In order to use the MicrosoftML and RevoScaleR libraries, you need an installation of Microsoft Machine Learning Server or Microsoft R Client.) A great way to see what MicrosoftML can do is to take a look at the on-line book Machine Learning with the MicrosoftML Package Package by Ali Zaidi.

The book includes worked examples on several topics:

• Exploratory data analysis and feature engineering
• Regression models
• Classification models for computer vision
• Convolutional neural networks for computer vision
• Natural language processing
• Transfer learning with pre-trained DNNs

The book is part of Ali's in-person workshop "Statistical Machine Learning with MicrosoftML", and you can find further materials including data and scripts at this Github repository. If you'd like to experience the workshop in person, Ali will be presenting it at the EARL Boston conference on November 1.

 

Github (Azure): Machine Learning with the MicrosoftML Package

Comments

[Javier Recasens]

I wonder if you can help me answer these two questions:
1.- Like RevoScaleR, MicrosoftML package can only use a maximum of two threads for analysis?
2.- Is it free to use at production level with Microsoft R Client?

[David Smith]

1. Both packages will use as many threads as available cores when used with Microsoft ML Server, and can also be run on distributed clusters and Spark.
2. Microsoft R Client is limited to 2 threads on a single machine. For production, use Microsoft ML Server instead.