| Student Projects |
|
Due to heavy demand for supervision, I now require students to meet certain grade standards and to have an adequate proficiency level in written English before I will agree to be their supervisor. Project students (except BTech) and MSc candidates should have a minimum GPA of 5.5 over their last year of study, preferably better. English requirements vary with what you want to do. English does not have to be your first language (as my many Chinese students, past and present, will happily confirm), but if writing a consistent document longer than 10 pages makes you go dizzy, then don't bother applying! By the way, by "writing" I mean writing in your own words, not in other people's words that you have cut and pasted from sources you found on the web (that's called plagiarism and if this is the only way you can think of putting a thesis or project report together, then again don't bother applying here - I smell plagiarism from miles away!)
Entropy estimation on time series data
Many complex processes produce time series data in the form of log entries, sensor data such as voltages, temperatures, pressures, fill levels, etc. or other observations. In a stable system, one expects this time series data to produce new patterns at a more or less constant rate. This rate can be described by the concept of entropy, which originates in thermodynamics but has reared its head in communication and information technology through the works of Shannon, Kolmogorov, Fisher, Lyapunov, Solomonov, Lempel and Ziv, Tsallis, and also more locally of Mark Titchener and myself - among many others. A stable system exhibits more or less constant entropy, whereas a system changing its overall dynamics will normally exhibit a change in entropy rate. Sounds theoretical? It is, to an extent, but there are many applications: a past PhD student of mine has demonstrated that it can be used to detect large-scale computer network events whose character is not known beforehand, Mark has shown that it can work well in a medical context, and now I'm looking for students willing to tackle other domains: industrial data, climate data, etc. to see whether we can demonstrate entropy effects there.There is more than one project in this.
Can we rely on timely in-order packet delivery on the Internet?
Somewhat related to the previous topic, with Internet real-time applications in mind, such as voice over IP (VoIP).In VoIP, the parties send streams of packets with voice data to each other. If too many packets fail to arrive or are unduly delayed or out of sequence (by having travelled along different routes across the Internet), then the receiving end will experience a deterioration in quality down to the point where VoIP breaks down altogether. With the rapid growth of the Internet, the number of routers and hence route alternatives grows, and so it's a little hard to tell whether the present situation (most packets arrive in time and sequence so VoIP works reasonably well) is actually sustainable.
The worrying bit is that chaos might actually creep up upon us, and that might threaten not just a big chunk of our international telephony, but also kill off a large number of the almost 30,000 jobs in the NZ call centre industry.
So we'd like to know whether we're heading for more stability in VoIP, or more chaos. We're currently in the process of putting experimental software together that we'll use to exchange predefined Internet traffic with a number of partner research groups overseas, so we can look at the shape in which traffic sent between far removed places arrives at the other end.
Again, there is more than one project in this.
The maximum feasible codeword length for T-code sets
This is a leftover from my own thesis - a little theorem I couldn't prove in the hurry. If you use T-Codes for variable-length coding of a source with N symbols, then I believe that (for binary channel alphabets), you don't need to consider any codewords that are longer than N-1 letters from the channel alphabet. OK, that's what I think, but I've started out with little more than a few handwaving arguments in my favour. I guess a proper proof of this little theorem could be worth a postgraduate project. In 2008, Aaron Gulliver and I proved some results on T-codes and cyclic equivalence classes, which could come in handy.Is this your topic? If it is, then you are probably interested in information theory and/or coding issues. You don't get all panicky when there's a proof with some mathematical rigour to be done. In other words, you're a theoretician. You like a challenge.
An ESD calculator for T-Code sets
T-Codes are known to be self-synchronizing. We also know how to calculate the expected synchronisation delay for a T-Code set that's used to encode a particular source. That is, we have the theory for it (which isn't entirely trivial) and it takes someone to put it into practice - a computer program, in fact. We've had a project student for this a long time ago - he didn't quite get there, so I'm prepared to call this a master's if it's done well and if that's what you're looking for.Is this your topic? You are not afraid to read yourself through a bunch of equations and through some theory to get started (with backup from me). You are interested in coding applications and you like to write pieces of software that do something mathematical with an application. You like to program in C and/or Matlab.
Your Own Topic
If there's something that you think fits within an area that I have an interest in and you'd like to do some work in that area, ask me. No one has a monopoly on good ideas, and if yours are workable and realistic then I'm happy to give it a go.
What do I expect from a project or thesis student?
I'm a very team-oriented person, and so the first and foremost expectation I have is good communication. It's only normal that you will encounter problems during your project. While I expect you to try and solve them first, I won't bite your head off if it's just too hard. However, you must keep me informed - if there is anything that impedes your progress, then we need to solve it before you're in hot water.Part of the good communication is a good report or thesis. I will ask you early on to start writing - reports and theses are not just a bloody chore that can be done in a lacklustre way. They are the central assessment tool - your grade and pass/fail depends on it. Don't get scared though - good writing is an extremely valuable skill to learn, and many people learn it during their project or thesis process. Also, don't be discouraged if your first language isn't English - believe it or not, but this is the easiest problem to fix!
Being team-oriented also means that you must be reliable. If you promise to do something, I expect you to do it. Don't promise me anything just to be polite. I prefer people who under-promise and over-deliver.
Any questions? E-mail me.