I've been working on my PhD since August 2009 at the University of Maryland, College Park.
Through May 2012, I was taking coursework and TA'ing in College Park. In June of that year I took a job in Rockville as a programmer. We moved to Rockville the next January, and it's proven hard for me to make steady progress on PhD work ever since.
So here I am in 2015, having done my dissertation proposal over 2 years ago (October 2013) and having no new publications to show for it.
I do have a good idea. At a high level, it's the application of a predictive machine learning technique to the selection of automatically generated model-based test cases for event-driven systems. No one has stolen my particular idea yet, or anything like that ... so I need to get it out there and graduate.
I have spent an extraordinary amount of time (way too much time, really) developing sophisticated automation for the grunt work of generating these test cases and "playing them", recording results, applying my technique and analyzing its impact. But the performance of my technique did not work out as I hoped, meaning that I'm basically doing something wrong.
I think the main thing I'm doing wrong is the identification of "features" for my machine learning algorithm. So I'm taking a step back to try to ID the important features first. Once I've ID'ed important features, I'll get back to predictions.
Now you're caught up.