Friday, December 31, 2010

New record for our data mining class

We got >100 people signed up for our 6 month series of data mining classes. The data mining class is the most popular because of the amount of work we put into redoing the assignments to make them all coding assignments vs. the proof type homeworks you get with the Stanford classes. CS229 didn't work with the after work audience. We still have a >50% drop out rate. This is the first time we are running it on Sat, it is too big for the Dojo now. We have to move. I sent out emails to Sebastian and Ari at Palantir. Going to try Sam at Facebook next. I want to get the students visible to potential hiring companies to put more pressure on them to be serious.

The Java Concurrency class is fun, working on building a server now. The code samples you get from an internet search on NIO servers is useless, all of them are wrong and none of them talk about the corner cases when a server would fail, like getting out of order packets, etc... Useless if the data doesn't get displayed correctly. We should turn this into a project class but if I am the only one doing the homework it may not be worth it. We can try turning it into a study group which meets once a month and demonstrate the projects we are doing. Then we can do accelerated lectures again using the JCIP book. The problem is there is a lot of low level stuff with locks you need before you can become a productive Server/Threads programmer. Lots of fun. You have to use a real server as a template like the Jetty server to understand the corner cases. Many changes in HTML5 like web sockets which should have an impact on evolving the next generation server.

We need a world class JS class to complement the server class. Going to talk to Dion/Ben and Mike Irish to see if we can develop a viable strategy for JS classes. The problem is I need someone to help run it. I can't handle the data mining, server, and JS classes at the same time. Nobody wants to do the dirty tactical work.

No comments:

Post a Comment