Jon Khanlian will discuss nets-vs-automata.net, a distributed browser-based study using tensorflow.js to compare neural networks' pattern recognition with cellular automata. The project maps a six-dimensional performance metric landscape over a thirteen-dimensional training parameter space, with hundreds of trillions of possible experiments. The talk covers when nets perform well and when they don't, using the nets-vs-automata analysis page. Attendees can read the related arXiv paper ahead of time (1809.02942). Speaker: Jon Khanlian, actuary, programmer, and filmmaker.
talk-data.com
J
Speaker
Jon Khanlian
1
talks
actuary, programmer, and filmmaker
Jon Khanlian is an actuary, programmer, and filmmaker interested in machine learning. He created related projects including scriptmonger.org and PathToTheCup.com. He has some experience with Haskell and Lisp.
Bio from: Jon Khanlian: (Neural) Nets vs. (Cellular) Automata
Filtering by:
Jon Khanlian: (Neural) Nets vs. (Cellular) Automata
×
Filter by Event / Source
Talks & appearances
Showing 1 of 1 activities