Microsoft today announced the release of a new open-source machine learning toolkit that goes by the name DMTK. The toolkit contains a framework for training models on multiple servers, a topic modeling algorithm, and a word-embedding algorithm for natural language processing. It’s all out on GitHub under an MIT (Massachusetts Institute of Technology) license.
Using these tools, developers can handle machine learning at scale with fewer servers, George Thomas Jr. of Microsoft Research wrote in a blog post today.
The timing of the release is uncanny. Google open-sourced a machine learning library called TensorFlow just three days ago. But there you go.
Facebook released open-source tools for deep learning, called Torch, earlier in the year.
All three of these companies are building up machine learning talent pools, and open-source code means that people outside their offices can collaborate and improve on it. These companies can even hire people from these communities.
Documentation for the new DMTK toolkit is here.