The first couple weeks of the Winter semester are over. Right now, I am taking one class on controlling non-linear systems using neural networks. Instead of a final exam, this class has a project. I spent a couple of weeks debating on what I wanted to do for my class project. I considered doing my class project on controlling a robot using neural networks and entering it in Cal Poly San Luis Obispo's Roborodentia contest. (This would be a cool way to show off Carnegie Mellon's Ph.D. program to potential applicants at Cal Poly SLO.) However, considering the amount of hardware development involved in building a robot in less than three months and the fact that I have other Ph.D. projects and quals coming up this Fall, I decided it would be better to do my class project on controlling a beamformer using neural networks. The beamformer problem is one of my Ph.D. projects, after all.

As I mentioned before, I have quals coming up later this Fall. For those of you who are not familiar with Carnegie Mellon ECE Department's Ph.D. quals process, I need to choose three published papers related to my Ph.D. work, write a four page document on these papers, and give a presentation in front of a panel of ECE professors (a.k.a. the qual committee). It is not sufficient for me to just summarize the three papers I chose, as I am expected to show my understanding of the topics by going beyond what was published. For example, I can find mistakes in the papers or explain how the authors could or might proceed with what they had previously developed. I had originally thought of choosing papers on solar antennas (my other Ph.D. project). However, through our weekly Ph.D. paper discussions, I soon realized that presenting papers on solar antennas would open me to questions by the qual committee in areas that I am not fully prepared to answer: Terahertz semiconductor physics, for example. Instead, I've decided to focus on papers related to areas I know well such as wireless communications, machine learning, neural network control, and beamforming using machine learning techniques. Of course, you may notice the tie-in to my beamformer Ph.D. project. After all, one (but no more than one) of the papers I choose for my quals can be a paper I wrote and published. I've written and submitted a conference paper on simulating a robot using machine learning techniques for my machine learning class last fall, and I plan on writing and submitting a conference paper on my Ph.D. and class project beamformer controlled by neural networks. After all, a great way of showing that I understand the topics I selected for my quals is to present a project that I developed.

On a bit of the fun side, and a shameless plug if you don't mind, I've entered the 2010 Dell Social Innovation Competition with my solar antenna Ph.D. project. You can read my entry and please vote (by March 1st) for it at http://tinyurl.com/solarantennas. The first round ends on March 1st. Not only do 50 entries selected by the judges move on to the semifinals, but the 10 entries with the most votes also become semifinalists. I think it would be really cool to win this competition, so I greatly appreciate your support. I'll let you know how I did after March 1st.