This post is sort of a hodge-podge of random bits, mostly just so I can post something new to the blog to prevent it from stagnating too long without anything new. The last semester has been far busier than anticipated. I got a paper accepted to ICC 2010, been a reviewer for a couple of conferences and journals, “started” the ACM student chapter at the University of Guelph and completed my course work requirements (hopefully).
The biggest lesson I’ve learned this semester is that the profs in undergrad were always right. Math is very important in CS. While it may not have been apparent at the time, (I always questioned the relevance of much of it since it was all presented very abstractly and disconnected from anything I was doing in undergrad) I now find some of my math skills to be my greatest weakness as a CS researcher. This is why this past semester I have gone out of my “comfort zone” and tried a couple of courses which I would generally shy away from. First is discrete optimization (Classifications of Optimization Problems, VRP, TSP, Bin-Packing, Cutting Stock, Simplex Method, PSO, Tabu Search, Branch & Cut etc). Second is a system performance and evaluation course (Queuing Theory, Markov models, Operational Analysis, etc). In each case, I have been pushed to the limits of what I know (and what I remember from my undergrad days). On the other hand, it has been easy to identify my weakest areas in my knowledge so that I can hopefully improve in them significantly before my quals in a couple of semesters. As an added benefit, perhaps because of the perspective I have gained as a grad student, I have started to realize the larger picture and see how many of the mathematical ideas can be applied to improve the state of the art in a particular problem area.
Lastly, I have been working at formulating my research and solidifying what it is I am working towards with this PhD. The very general description is “heterogeneous wireless networks”. Within this area, I am interested in particular type of heterogeneous wireless network. It is made up of a wireless mesh network as the backbone. Instead of traditional client nodes such as laptops, pdas etc, the clients are actually other types of wireless networks, for example: 802.11 WLAN, RFID, 3g Mobile etc. There are many problems when this type of network is considered. One is the gateway placement problem (where to place gateways between the types of networks, how many etc.). Also in each sub-network, there may be a particular set of parameters which cause the sub-network to perform very well. However when the sub-networks are joined, this set of parameters may not be optimal for the entire network as a whole. The problem then is how to optimize these sets of parameters with respect to the entire network.