Super hard data challenge

[challenaccepted]

No description provided.

[nottombrown]

Good work with this, CA. You had the second highest accuracy with 294 correct predictions. Email me at tom@joingrouper and we have a sweet prize.

[challenaccepted]

Sorry for the nasty code; I didn't have enough time to beautify before the
deadline, and I figured a sloppy PR was better than none.

I thought that was a really super hard data challenge--signal-to-noise
quite low in the training data. My training accuracy never got above 60%,
and in the end, I abandoned all fancy methods in favor of a stupid linear
separator.

I worked on it a bit more, and my accuracy is up a lot higher now. If
you're interested in the results, I'm happy to send them over.

Which approach did the winner take?

On Tue, Dec 17, 2013 at 7:55 PM, Tom Brown notifications@github.com wrote:

Good work with this, CA. You had the second highest accuracy with 294
correct predictions. Email me at tom@joingrouper and we have a sweet
prize.

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/10#issuecomment-30807188
.