Automated Machine Learning (AutoML)
I'm hearing more about AutoML lately.
The basic idea of AutoML is to automatically set up a machine learning pipeline from data to deployment.
Another intern of NS will focus on AutoML, so hopefully I'll learn things from him.
My first thoughts: if your only tool is a hammer, everything starts looking like nails.
I'm quite sceptical of this idea, because one of the things where ML and data science practitioners really add their value, I think, is determining which machine learning approaches are suitable for a particular type of problem and type of data. To determine that, you may need a lot of data exploration as well as understanding business problems. I don't really see how you can automate that side of machine learning. I don't even think the main challenge in applying machine learning has much to do with techniques; I think it may at least for 50% be understanding business and where machine learning really may add value (and where not).