Machine Learning Lecture #5 Discussion

[Dave Colwell]

John, Grant, and I met this morning to discuss lecture #5 from the
Machine Learning series. John helped to clarify some topics related
to Ng's derivation of a Sigmoid curve from a linear plot of example
data with Gaussian classifications. We also talked through the Bayes
naive classifier, which presents a nice segue back into the
Programming Collective Intelligence book. John proposed that we work
through the chapter, including pages 122-123 on Bayes filtering.

For Friday's Geek Book Club meeting, one topic on the table is a
continued discussion of P vs NP and the zero-knowledge proof.

[smantix]

At the meeting, I recalled that Paul Graham had written an article on
Bayesian filtering for spam, printed in Hackers & Painters. It is
available online as "A Plan for Spam" at http://www.paulgraham.com/spam.html.
There's also a sequel, "Better Baysian Filtering" at
http://www.paulgraham.com/better.html.