0

Graphical Models

I am finally at the point where I love the research I am doing, namely Bayesian Networks and Graphical Models.  This particularly stood out to me when I read it.

“Graphical models are a marriage between probability theory and graph theory. They provide a natural tool for dealing with two problems that occur throughout applied mathematics and engineering — uncertainty and complexity — and in particular they are playing an increasingly important role in the design and analysis of machine learning algorithms. Fundamental to the idea of a graphical model is the notion of modularity — a complex system is built by combining simpler parts. Probability theory provides the glue whereby the parts are combined, ensuring that the system as a whole is consistent, and providing ways to interface models to data. The graph theoretic side of graphical models provides both an intuitively appealing interface by which humans can model highly-interacting sets of variables as well as a data structure that lends itself naturally to the design of efficient general-purpose algorithms.

Many of the classical multivariate probabalistic systems studied in fields such as statistics, systems engineering, information theory, pattern recognition and statistical mechanics are special cases of the general graphical model formalism — examples include mixture models, factor analysis, hidden Markov models, Kalman filters and Ising models. The graphical model framework provides a way to view all of these systems as instances of a common underlying formalism. This view has many advantages — in particular, specialised techniques that have been developed in one field can be transferred between research communities and exploited more widely. Moreover, the graphical model formalism provides a natural framework for the design of new systems.”

Michael Jordan, 1998

0

Progress

Hurray! Dann and Daniel arrived last Sunday and it’s awesome to have them around. I couldn’t sleep at all on Saturday night due to the excitement of it all. This showed when we all arrived home from the airport and crashed for 3 hours!

The ride on the Airport Express to pick up my boys!

The ride on the Airport Express to pick up my boys!

Daniel was apparently behaved on both flights – they flew Brisbane to Singapore then Singapore to HK. Though apparently he woke up 3 hours out of Singapore and refused to go back to sleep (even though it was about 2am). Poor Dann. Though all Daniel wanted to do was watch TV shows. It’s kind of good that he can enjoy the onboard entertainment these days!

First evening in HK for Dann and Daniel - checking out the Symphony of lights at TST Boulevard.

First evening in HK for Dann and Daniel – checking out the Symphony of lights at TST Boulevard.

So the first week with them here consisted of showing them the lay of the land: where good restaurants are (of which there are plenty!), where the good playgrounds are located (there are no bad ones that I’ve found), the nearest supermarket (about 4 doors down) and getting them oriented with the place. Of course, there was the obligatory trip to the mobile phone shop to get Dann connected to the interweb. Pretty much everything that took me about 10 days to figure out, they got the crash course in a day and a half, and every afternoon after I finished up at uni for the day.

School work everyday for Daniel!

School work everyday for Daniel!

For the most part, we’ve settled into a routine. Dann and Daniel walk me to Uni (it’s awesome living 10 mins walk away instead of the usual >1 hour drive commute from Ipswich to Brisbane) and we have a morning coffee before I go to work. Then they both go for a longer walk either to a playground to run around, or to visit a local site or something. Last week they went to the Hong Kong Science Museum. Daniel absolutely loved it, which was awesome. He raved about it for days and days and wants to go back again and again. At a cost of $25 HKD for Dann, or around $3.50 AUD (Daniel is free) it’s pretty good value for money!

Afternoons at Kowloon Park!

Afternoons at Kowloon Park!

After this, they usually head off to have morning tea in town or in the apartment where they then do Daniel’s school work for the day. We were initially worried about the stuff that Daniel would miss at pre-school by being here in HK, but because his pre-school is awesome (I love them and cannot rave about them enough), his teachers put together a heap of worksheets that he can do while he is here. We also have a book that Daniel draws in and sticks all the paper souvenir things he has (ticket butts etc) so that he can take these home to remember his time in HK.  In order to keep him and his friends at pre-school connected, I’ve been sending them photos of the things he’s been doing and also a short newsletter-type thing that tells his friends back in Australia all the things he’s been doing in HK.

Playground time.

Playground time.

Following school work time is lunch, then nap. Dann’s been taking this time, as well as the afternoon time after naps when Daniel plays with his toys/on the iPad/watches telly, to do the documentation that he needs to write for the new BMath degree at QUT.  He finally finished this yesterday (it was due today), but he had to work through last weekend to finish it. So he’s more relaxed today, and next weekend we can all do something together!

The time after I finish up at Uni are usually spent together at the park. We’re quite frequent visitors of the playground at Kowloon Park these days, as well as the other playgrounds around TST! It’s really, really nice to have them both with me here.

Last weekend (while Dann was finishing the BMath documentation), Daniel and I went to the Hong Kong Space Museum and Middle Road playground. This was a great way to spend a Saturday. On Sunday, we made our way to Noah’s Ark in the New Territories. Again, it was a nice day out for Daniel and me.

Cooking up dinner in our little apartment.

Cooking up dinner in our little apartment.

Work-wise, I’m still thinking of way in which I can incorporate Dr Tam’s work into my wayfinding BN. It’s hard as her data is more on levels of service stuff, but I have a couple of ideas which will hopefully pan out. I’m meeting with her and Prof Lam again on Wed morning.  In the meantime, I’ve been surprisingly productive doing other things.

I’ve finished all I can do on a paper that the Complex Systems team I’m in are working on. It’s on visualisation of BN model results and is shaping up to be a really nice paper.  I’ve also started on my BN methods paper, which is a good thing. I have a fairly solid idea of where this paper is headed (finally), so it’s nice to be able to make some progress on it – even if this progress is just putting my references in BibDesk and summarising the relevant papers to which I’ll be referring!

Anyway. I’ll end this post on some dot points that I’m too tired to elaborate upon (I really haven’t been sleeping well at all and it’s all starting to catch up with me):

  • While I love having Dann and Daniel here, I still think that this whole thing has been quite selfish on my part. Uprooting everyone and so forth. I know it’s a great experience for as all, and as Dann says ‘It’s what academics do’, I still feel guilty about it. Particularly when Daniel gets upset and says ‘I miss home’. That gets me every time.

    My little boy. I love him so.

  • I still have no freaking idea about what I’m doing. Apparently this is a common feeling among academics who aren’t completely up their own arses. Or so I’ve been told. I’ve been feeling this way since I started this degree and it hasn’t stopped. I doubt it will.
  • Looking beyond PhD completion, I have no idea where I’m headed. I like academia, but I’ve also seen the bad side of it too (in work and personal lives of others). I guess that’s the same everywhere though.  The question is: Can I survive in the real world, and would the real world even want me?! For the first time in years, I actually have nothing in place after I submit my thesis in a few months’ time. I’m at the point of trying to balance my time doing my work and applying for post-doc jobs etc.

That’s it for now. More to come once I get some much needed sleep. Though I’m beginning to think that’s not going to happen until after I hand in this thesis.

Le sigh, le sigh.

Afternoons spent together at Kowloon Park!

Afternoons spent together at Kowloon Park!

0

First week – Self doubt is a bitch.

Phew, well I’m glad that’s over.

I had my first meeting with Prof Lam at 10am last Monday. As per normal, my irrational fear and hate of being tardy made it so that I was outside is office almost an hour before I had to be. Some things never change. It did give me a chance to wander around the campus. Size-wise, the campus isn’t huge. It is so very very clean though!

HK Poly USo the plan for my 3 month study here is to modify my current wayfinding model and to apply it to HKIA to assess wayfinding effectiveness there. I contacted Prof Lam at HK PolyU specifically because he, and a student of his, had some some work on wayfinding at HKIA as well as Levels of Service at the same airport. So, I’m talking away with him about what my model was and how it was different (novelty etc being important) and what it contributes to the literature and wayfinding research in general. This all seems ok, though he’s not entirely convinced that a Bayesian Network (BNs) is a good idea for this. Then we get onto how I quantified the model. He doesn’t think I have enough data. This is a problem that he keeps going back to over and over again. I’m trying to make the point that BNs don’t necessarily always need survey data to use for quantification. That’s one of the strengths of this modelling technique. You can use data sourced from different places: survey, experts, literature review, studies. Anyway, this is a sticking point.

After my meeting and (quiet) meltdown. I did what any self respecting woman would do. I bought a new (enormous) handbag.

After my meeting and (quiet) meltdown. I did what any self respecting woman would do. I bought a new (enormous) handbag.

I explain the rest of my model and try to point out that the previous mathematical models, while good, only take into account the airport environment when assessing wayfinding. By its very definition, wayfinding is a combination of personal factors (your cognitive skills, ability to form a cognitive map, your spatial skills etc) and environmental factors (signs, landmarks etc) that come together (things process in your brain) and you get from point A to point B. The wayfinding BN we are proposing takes into account both of these types of factors. Anyway, he’s still not convinced and is trying to say in a nice way that my model doesn’t prove anything mathematically, the results aren’t reproduceable, and (this is what I’m hearing) I have nothing.

Cue holding back tears and trying not to cry. It’s tough. I’m fairly sure my supervisor would have told me if what I have been doing for the past 2 and a bit years is nonsense. I’m fairly sure she would have pulled me up and said ‘Whoa – let’s do something else. This line of research is not going to get you your PhD. It’s not novel, there’s no new methodologies, the application isn’t new.’. But she never said any of those things – I’m trusting that she knows what she’s doing. Given that she’s successfully supervised umpteen students, and has had a squillion research grants. So, trying to keep this in mind, I keep talking with Prof Lam. (I’m making him sound mean here, but he really is very very nice).

Construction of the new School of Design building at HK PolyU. This building will be awesome when it's finished.

Construction of the new School of Design building at HK PolyU. This building will be awesome when it’s finished.

We eventually work out a plan that involves me working with one of his previous PhD students (she now works at HK CityU) ad we’ll try to expand the wayfinding BN to include more factors from the Visibility Index models of wayfinding. I’m happy with doing this as I think that research is good, and it will give the wayfinding BN more rigour. we meet with Dr Tam later than afternoon and all agree to this course of action.

Now I’ve got 2 weeks to figure out some ideas about how to do this. I have some fairly solid ideas that can also use some data that Dr Tam has already collected from her PhD. That way we make use of everything we have.

Of course, after the end of my first meeting I wrote to Kerrie about it. She made some very good points. Yes – people aren’t going to always like your work. But take that on board and see what they can offer to make it better. That will make me and my research better in the long run. Also, it’s good that they don’t agree with it. That’s fine. It will give me work to do. Dann said the same thing – after talking me down from my ‘What the hell am I doing here?! I shouldn’t have done this. This was such a bad idea. I need to go home NOW.’ melodramallama episode.

They are both, of course, right. I need to get a thicker skin. I need to stop being such a pansy and so freaking sensitive.

In the wise, wise words of Kerrie, “Feeling like crap is totally expected at this stage! You shouldn’t, but I can imagine that you do. But believe in yourself and your research!”. I’m trying to do that. It’s hard. I don’t know how other people do it. But I’ll get there.

0

Conference Talkies & other tidbits

Today I received notification that my CTAC abstract was accepted. Yay! Another publication before the end of the year. This follows on from the acceptance of my paper with Transportation Reviews.

I may just get this degree on time! 😀

In other news, I downloaded Bayesian Reasoning & Machine Learning by David Barber today. It would be awesome if I managed to get through the 600-odd pages of it. If I get through that and Koller and Friedman’s Probabilistic Graphical Models: Principles & Techniques (which I ordered last week), I’ll be quite chuffed.

I’m trying to also be brave and contact overseas Bayesian Network and Probabilistic Graphical Model researchers to see if I can talk to them/their group sometime next year. So far, I’ve only managed to look up people I’d like to meet. It’s quite intimidating.