Machine Learning: Learn Now

July 18, 2016

If you want the basics taught in most universities, you can start with the papers listed at this link. If you come away from these write ups with some questions, you can refresh your knowledge of Bayesian machine learning in a paper of the same name. To get a sense of some limitations of the much-hyped “new” approach to smart software, check out this sort of slideshow, sort of lecture called “What’s Wrong with Deep Learning?” Balanced views are difficult to track down. There are the cheerleaders, and then there are some implementers. A representative example of cheerleaders are the ad hoc team of Google, Microsoft, and some startups, research computing outfits, and lots of academics. The doubters are old people like myself who have had to deal with the interesting “drift” which can creep into deep learning systems. What’s drift, you may ask? Well, you expect one thing and get another. No human knows why. That’s drift.

Stephen E Arnold, July 18, 2016

Comments

Comments are closed.

  • Archives

  • Recent Posts

  • Meta