Machine Learning Resources

Posted by admin on April 30th, 2016 filed in Development

Suddenly everyone wants to solve their IT problems by throwing “machine learning” at it. (Well, I want to do that, too…) Perfect time to read up on what “ML” means!

Here’s an intro that cover what machine learning can and cannot do, what are artificial neural networks, what calculus and statistics knowledge is needed to not embarress yourself, =-< which langauges and libraries are good for rapid prototyping, and what large-scale tools are available for us to set up a prototype. Udacity/Roojoom: Introduction to Machine Learning

My main take-away is: Powerful tool, powerful garbage in, powerful garbage out.

Machine Learning will not magically do the thinking for you. It will not magically find the deeper truth in reality by looking at your harddrive. It will take your hastily annotated, roughly digitized data, and it will teach itself to magnify any accidental random pattern in this data, non-transparently, a millionfold. If you catch yourself thinking “Why that’s an odd result, but I guess the computer knows better than me” — dig out your knowledge of statistics, use your scepticism w.r.t. the training data, put the results in perspective, and set the record straight with your boss, before he uses the results to decide whom to imprison and whom to hire, please.

Comments are closed.