Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Show HN: A machine learning Rails app to predict your favorite color
8 points by eloisius on Dec 15, 2011 | hide | past | favorite | 4 comments
I'm nearing completion of the machine learning class offered online by Stanford (Prof. Andrew Ng is awesome).

I wanted to try out some of the concepts I learned so I made up this little Rails app. I doubt if you can actually predict someone's favorite color based on a few preference, but we'll see how it looks after a few (thousand?) examples. I'm still working on it, but it's in a working state and I'm collecting all the data I want, so I can always rehash it later to tweak the algo.

Any suggestions for a more interesting set of data to form predictions on?

http://colorsvm.herokuapp.com/

https://github.com/zacstewart/colorsvm



Stumbled on a bit of a usability problem. On the results page I clicked "correct" without thinking, because it got the answer correct. I then read a bit more and realised what the button was only for when it gets the answer wrong.

There's no way to acknowledge a correct answer, which is a piece of feedback you might not want to miss out on.

Or maybe I just think about things in a weird way and should think before I start pressing buttons.


You're right. Using the word correct does create some ambiguity. Recording a "correct" answer probably wouldn't provide me with a lot of data. It would be doing the same thing as "correcting" it with the same answer that it predicting. Basically, just further reinforce whatever made it predict right in the first place.


And then perfectly timed Redis To Go outage knocks it offline...

Should be back in a minute.


Enjoyed reading through your project's commits -- thanks for sharing.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: