[Noisebridge-announce] Machine Learning Meetup: making our neuron smarter!

[Josh Myer]

Wednesday nights at 8PM, we do a session on machine learning at 83c.
Last week, we had a hands-on workshop, and by the end of the night,
everyone left with their very own baby neuron.

This week, we'll be taking these baby neurons and making them a little
bit smarter.  By generalizing, we'll set ourselves up to build a
proper neural network from scratch this month, and get a better
understanding of how simple interactions can yield complex results.

You can check out our results at
https://www.noisebridge.net/wiki/Machine_Learning_Meetup_Notes:_2009-03-11

As with last session, there's a small bit of homework: glance through
your code!  We'll quickly review the ideas at the beginning of the
night, then move on to the new stuff.  I might need to assign new
people to work with people who were here last week at first, so be
ready to explain how a perceptron evaluates input and a quick overview
of the training process.  (And if you weren't here last time, have a
look at that wiki page: there are perceptrons in python, ruby, matlab,
mathematica, C, and, yes, even LISP.  There's also a tutorial linked
from the notes page.)

The goals for this week are:
 * talk about overfitting in the training set
 * refactor to continuous inputs and outputs
 * (hopefully) change from a linear basis to sigmoids

I'm looking forward to seeing everyone there!
-- 
Josh Myer   650.248.3796
  josh at joshisanerd.com