Cooperative Bug Isolation
It is standard engineering practice to monitor and measure the structures we build as they are being used, but software is an exception. Most software, once deployed to a user, is not monitored at all or is monitored only according to an engineer's best guess about what may be important to observe. As a result, we believe a great deal of useful information about the behavior of a program, information that is computed for free by the program's user community, is simply lost.
This talk presents techniques for the systematic monitoring of thousands to millions of distributed program executions. We discuss how to exploit this information in a particular application: Using partial information gathered from many program executions to automatically isolate the causes of bugs.
Statistical debugging using diagnostic and predicive inference. Very cool stuff! [more here later]
He's working with Michael Jordan at UCB, too (that's Michael I. Jordan of Jordan networks and Jordan-Jacobs Hierarchical Mixtures of Experts, the computational neuroscience, graphical models, and variational methods pioneer; not Michael J. Jordan, the retired NBA superstar). I had a few questions about machine learning in their apparently quite successful projects; he referred me to Michael for a couple of them, and also invited me to e-mail him about them. I mentioned that Michael made a very cogent comment at UAI-1997 that turned out to be important in my dissertation research, but forgot to ask him to thank him for me. Oh, well: I'll get my chance in Edinburgh come July, with high probability.
Cross-posted to grantwriting (which, BTW, is now a happening community that you should read!):
Is anyone familiar with the U.S. Department of Defense Small Business Innovation Research / Small business Technology Transfer (SBIR/STTR) programs?
I've been aware of SBIRs (pronounced "sibbers") for some time, but have never applied for one before. Similarly with STTRs, though I have talked with groups about putting one in.