Cortical Algorithm at PyGTA Tonight (Book precis and implementation discussion)

Since I didn't get anyone else standing up with a "deep" topic for tonight, I'm proposing that we look at the cortical algorithm put forward by Hawkins in On Intelligence, and then see if we can hash out how you'd go about creating such a thing in Python (efficiently enough to make it useful).

We'll be at Linux Caffe, as usual.

[Update] Grr, turns out the algorithm Hawkins described is patented. At least, some aspect of their implementation in software is. I'd have to look into whether you can patent direct mapping of biological processes, but I'm guessing at least in the States and Canada you can (you can patent anything here, seemingly). Which seems like a pretty good way to just kill research in the area. Oh well, guess we'll discuss the general case more than their particular algorithm.


  1. James Tauber

    James Tauber on 01/16/2008 12:26 p.m. #

    I didn't find enough of an algorithm in On Intelligence itself; I had to look at their published papers.<br />
    <br />
    Of course, Numenta has already implemented the algorithm in Python.<br />
    <br />
    But in essence, the algorithm in the book and related papers is just a Bayesian belief network with Pearl's propagation algorithm.<br />
    <br />
    You could implement that in Python unencumbered; I've toyed with doing that since I first read Hawkins' book.

  2. Mike Fletcher

    Mike Fletcher on 01/21/2008 9:15 a.m. #

    I've quickly read through a paper on the propagation network. Will need to sit down and think it over to see what the code would look like for it. Thanks for the summary and pointers!

Comments are closed.


Pingbacks are closed.