Our best work goes unnoticed, Machine Learning at scale
The recent advances in language and speech recognition are more down to the amount of data that can be reasoned against than improvements to the underlying algorithms. Also if a model is really successful then it likely to be widely used then it needs to run at scale as well. Finally models need to be operationalised quickly and be capable of being retrained programmatically to keep them fresh as new data is made available.
So the question is how to scale Machine Learning? Not only that, machine learning needs data scientists to work yet much of the work around operationalization and accessing high volumes of data is not natural to these rare and expensive experts. What they need is a workbench to do all of this using familiar tools and libraries, where they can clean and wrangle data and then deploy all their work to some engine to run that at scale. It is that need that is addressed by the new Azure Machine Learning Workbench and that’s what we’ll explore in this session.
It’s only just been launched in Public Preview so this is probably the first time you’ll get a chance to see it an action, although for the most part there isn’t much to see - it is a command line interface as that is the way most people work in data science!