Show notes
Jon and Catie walk through US Patent 9,406,017, Google's patent on dropout, a way of training a neural network by switching off part of it at random. Starting from what a neural network is and why one can memorize instead of learn, they build up to what claim one covers, what it leaves out, how the patent family grew in a chain of continuations, and four drafting lessons for anyone patenting artificial intelligence.
Sources:
US 9,406,017 B2, System and method for addressing overfitting in a neural network: https://patents.google.com/patent/US9406017B2/en
Claims of US 9,406,017…
Website: share.transistor.fm

