Netflix has changed the way people rent movies. When they started offering DVDs through the mail with no late fees, I thought it couldn’t get any simpler. Their addition of content streamed directly to your computer seemed like a nice feature, but it didn’t compare to the comfort of watching from your living room.

Recently, Netflix has been finding ways to connect home televisions to their streaming service. Newer game consoles, Blu-ray players, and televisions are beginning to come with this feature pre-installed. What if you don’t have any of these cutting-edge devices? There’s a solution for you too.

A company called Roku produces a small device which connects to your home network and the back of your TV. After some minimal setup (about two minutes, in my experience) you have access to all 50,000 titles in the Netflix streaming library. The Roku player supports both wired and wireless connection to the Internet. Composite video is the standard output connection, but HDMI is also available.

Kristen and I have had our Roku for about a week now, and we really enjoy it. With thousands options at our fingertips, we’ve been able to catch up on some television shows we’ve been waiting to see. Even though our Clearwire Internet connection is pretty slow and spotty, the Roku player is still able to serve up video with medium to medium-high quality.

I would easily choose a Roku and Netflix combination over a cable or satellite subscription with hundreds of channels of worthless programming. The value lies in the ability to watch what you want, when you want, without the distraction of advertisements.

The Netflix Prize class project finished this morning at 12AM. My team won with an RMSE of 0.9028! Here is a copy of our report, detailing our approach and our solution. If you follow the instructions in Appendix B, you should be able to reproduce our best result. In the future, I plan to clean up that script a little bit and post it to the Netflix Prize forum.

My Netflix Prize team has improved its RMSE to 0.9028. We are very excited at our progress. Our algorithm is now 5.11% better than Cinematch. Half way to $1,000,000!

For my machine learning class, the contest ends Wednesday. On Thursday I’ll post my team’s final report, along with instructions on how to reproduce our methods.

My team has a new best RMSE score for the Netflix Prize project, 0.9066. After my last update, we were overtaken by another team in the class. With this submission we are back to first place in the class and have jumped up to 449th place in the world. The class competition finishes on December 17th, so we’re closing in on the end. I’ll post my team’s final report once the dust settles.

Lately, a lot of my time has been dedicated to the Netflix Prize project. It’s part of my CptS 570 Machine Learning class, and the goal is to improve Netflix’s movie recommendation tool. I’ll elaborate on the details of our algorithm later, but here is my team’s current status. We are down to a Root Mean Squared Error (RMSE) of 0.9261, which is 2.66% better than what Netflix can do. In the overall worldwide competition, our team is in 1005th place. For now, we are also the leading team in the class. Check out NetflixPrize.com and the class project page for more details.