brisbane public transport routes

Yesterday I caught the 104 bus from the Corinda Railway station to the PA hospital. Apart from the Great Circle 598 and 599 routes and the various routes servicing the University of Queensland in St Lucia from the western suburbs, I think this must be one of the very few buses that doesn't run on an into-the-city/out-of-the-city axis. But it only runs Monday to Friday. What a shame. I suppose in 2007 with the Green Bridge opening across the river, that will be another way to get from Corinda to UQ.


social networking games

Linkedin.com is fun. Here is my public URL. I am connected to 13 people and waiting on another 11 invitations. Instead of guessing people's names I just went through all of my email and took everything with an @ sign, then uploaded it to the contacts list on linkedin. Now it says "You have 8450 contacts and 120 are already users". I've had some contact with 16 of those 120. It's interesting seeing what people who have some connection with me are doing.

I would like to get in contact with friends on Myspace but I don't seem to know many. Perhaps I'm too old (turned 29 3/4 yesterday) and Alexa statistics merely reflect its huge popularity among teenagers.


ghormeh sabzi

Last night I made my first ghormeh sabzi. Even though I used chicken instead of lamb/beef, lentils instead of kidney beans, and lemon juice instead of the usually recommended limes, it still tasted really good. The recipe I followed was from Forough Hekmat's "Art of Persian Cooking".


social networking

I was speaking to Jenny on the phone the other day and she said she received a message on hi5 from a mutual friend, Nami, after a prolonged period of no contact. So I decided to do the rounds of invitations again on the services I'm a member on. hi5, friendster and myspace have a useful feature where you can invite your MSN and Yahoo friends automatically, but it seems to be broken on myspace today.

From this list I am a member on Friendster, Hi5, Myspace, Orkut, Wayn, Ryze, Tribe.net, LinkedIn, and an unlisted one, Gazzag. That last one is Brazilian and looked a lot like Orkut last time I checked.

My favourite is still Orkut, I have more friends there than on any other service (107).


these people

Googling "only thing arabs understand", "only things these people understand", and "only thing they understand" arab gives predictable results. I wish I could google in Arabic, so I could see if Arabs say similar things about Israelis.

In a story from news.com.au Israel requested jet fuel from the USA.

The Pentagon said the sale would help "improve the security of a friendly country which has been, and continues to be, an important force for stability and economic progress in the Middle East."

Surely no-one from the Pentagon was able to say that with a straight face?



At trivia tonight we won the $500 grand final prize (taking into account the scores for the last 9 rounds). It was an interesting format that I hadn't participated in before and which John explained to me beforehand. The host distributes cards marked A, B, C, and D to the top four teams (plus a random team) and asks a multiple-choice question, repeating it once, and then on the count of three a person from each team holds up their answer. The last question was "which country does not border on Iran" -- A Afghanistan, B India, C Turkmenistan, D Turkey. And somehow the team which almost always came first put A, I think, and was eliminated. The other remaining team held up C. Clearly even our most competent opposition are just a bunch of bogans, at least when it comes to geography. (I can't explain how we scored 88 and they scored 100 tonight). We can't rest on our laurels though -- got to keep our minds sharp and our team name fresh!

move 15 in Kasparov vs World

So I ran a few programs on this position.

Spike and Toga II like ...Ra8 and Rybka likes ...b5. "The World" played ...Ra8 but Kasparov feared ...b5.

Spike 1.1

info depth 24 score cp -32 pv h8a8 c1e3 e7e6 e3b6 f6d5 c3d5 c4d5 d1d5 e6d5 f1b1
a8a4 b6e3 d7c7 b1d1 g7b2 a1b1 b2d4 e3d4 c6d4 b1c1 c7d7 d1d2 d4e6 d2d5 a4a2 c1d1
a2a6 d5d3 hashfull 1000 nps 527191 time 297483992 nodes 156831064410 tbhits 0

Spike 1.2

info depth 24 score cp -15 pv h8a8 c1e3 e7e6 e3b6 f6d5 c3d5 c4d5 d1d5 e6d5 f1b1
a8a4 b6e3 d7c8 g2g3 c8c7 g1f1 d5d4 b2b3 d4e3 b3a4 e3f2 a2a3 g7a1 b1a1 f7f5 f1f2
d6d5 hashfull 1000 nps 564996 time 209331429 nodes 118271594633 tbhits 0

Toga II 1.2.1a

info multipv 1 depth 25 seldepth 82 score cp -12 time 342210312 nodes 180871282838 pv h8a8 c1e3 e7e6 e3b6 f6d5 c3d5 c4d5 d1d5 e6d5 f1d1 g7b2 a1b1 b2c3 d1d5 a8a2 b1d1 c3e5 b6e3 a2a3 d5b5 d7c8 d1b1 c6d8 g2g3 h7h5 e3b6 d8c6 f2f4 e5f6 b6f2 c6a5 g1g2 a3a2

Rybka 1.0 beta

info depth 23 score cp 0 time 351281642 nodes 1461490788 nps 4260 pv b6b5 f1e1 h8c8

rybka; kangaroos

I had lunch with Alejandro today. He wrote the smallpotato chess program. It takes a lot of effort to write a good chess program. We talked about the programs that came out of nowhere. First there was Ruffian by Per-ola Valfridsson of Schweden. He was not a participant in chess groups or discussion boards and then one day he released Ruffian which just beat all the other programs available. It was extraordinary. Alejandro told me that the same thing happened with Fruit, for which the source code of one version is available. He said it doesn't do anything special, just does the usual stuff very well. Finally in December 2005, the international chess master Vasik Rajlich came out with Rybka which is clearly the best program anyone has written up until now. He must have some new idea which ensures its superiority, because there is no way to "cheat" in computer chess. You can tune your program against some other programs to ensure good results, but you can't do it against all of them. It's impressive, because usually the best chess players aren't necessarily the best chess programmers.

On Saturday I was driving to Venman Reserve when a kangaroo jumped out in front of the car. It was rather frightening because there is no effective action possible. ABS brakes would have helped, though, so my next car will surely have those.


silly nationalism

On the night of July 2, 2000 I ascended the Eiffel Tower in France. Everything was quiet and peaceful. Suddenly there was lots of noise from down below! I descended again and Paris was in the middle of a big celebration. France had just beaten Italy and won the 2000 European Football Championship. Will there be similar celebrations after this year's World Cup final? As an Australian, I hope to see Italy defeated :-)

I watched all the Australia matches because of the feeling of euphoria when we scored. I admit though, it makes sense to be proud of your country when it is small, but I find the nationalism of large countries kind of amusing. If someone says they are proud to be Chinese or Indian, particularly, I think "you are hardly special, coming from a country that makes up 1/5 or 1/6 of the world's population. Wouldn't it be more sensible to be proud of belonging to a smaller sub-group?". Perhaps what they are really proud of is some kind of values the country is perceived to have, like Indian parents in the US celebrating their children's success in the national spelling bee. That would make sense.


long dongs

People have known about the Taepodong since August 1994 and the Taepodong 2 since June 1995. After 11 years North Koreans still can't get the thing to fly. Doesn't seem to be much of a threat.


I like the concept of Digital Object Identifiers. I need to make sure people can find my papers easily, and the idea of a permanent URL is exciting. Here are all my published papers with DOIs (nine out of 20, soon to be 10 out of 21).

A mixture model-based approach to the clustering of microarray expression data

A simple implementation of a normal mixture approach to differential gene expression in multiclass microarrays

Modelling high-dimensional data by mixtures of factor analyzers

Using mixture models to detect differentially expressed genes

Steiner trades that give rise to completely decomposable Latin interchanges

Cluster Analysis of High-Dimensional Data: A Case Study

Application of Mixture Models to Detect Differentially Expressed Genes

Clustering objects on subsets of attributes (with discussion)

A new bound on the size of the largest critical set in a Latin square

I am able to link directly to other papers I wrote, but the URLs are not permanent.

Closing a gap in the spectrum of critical sets

Robust cluster analysis via mixture models

And two more available through the arxiv:

The size of the smallest uniquely completable set in order 8 Latin squares

Critical sets in the elementary abelian 2- and 3- groups

OS X 10.4.7

Apple released an update for Mac OS X, 10.4.7. I installed it without thinking too much. Unfortunately, I have an April 2004 Powerbook and need to put a PCI card in the PCI slot to get any USB 2.0 slots. This update just doesn't work with the USB adapter in the PCI slot at all - things are very slow and upon waking from sleep, the mouse pointer works but I just get the spinning beach ball of death. I'm not happy and lots of other people have the same problem. Fortunately it seems possible to downgrade that part of the update.

EDIT 18 July: "the same problem" link disappeared. So I am copying the fix from here.


Make sure you have a good backup. If this fails for you, you will need to Archive & Install.

I was having the same issue with the computer not waking correctly from sleep after the X.4.7 update discussed in this thread. The problem had to do with having a USB 2.0 PCI card installed.

To fix it, I replaced the /System/Library/Extensions/IOUSBFamily.kext file from the 10.4.7 installation (2.5.6) with the 10.4.6 version (2.2.8).
If you're already on 10.4.7, you'll have to revert to 10.4.6 or use an app like Pacifist to get the 10.4.6 Extensions folder and copy it to another location on disk. Then update to 10.4.7 and after restarting replace the IOUSBFamily.kext installed with 10.4.7 with the copy from 10.4.6, repair permissions, and restart.

So far everything seems to be back to normal and I can put my G4 to sleep and have it wake up ready to go again.It worked for me, hopefully it'll help others having sleep issues.

NOTE: I don't know yet if this will cause conflicts later, but again, so far so good. This was a last resort effort after normal maintenance and disk repair, resetting NVRAM, verifying that it was not account specific (it failed to wake from sleep at the login screen after a restart), and several attempts at Archive&Install, Combo Update from 10.4.0 and 10.4.6, and Delta update from 10.4.6

weekend work

I was at the Gold Coast over Easter and played a game against someone in Charlie's Cafe in Cavill Mall. I lost but I found the opening interesting. As White I played 1. e4 e5 2. f4 d5 (Falkbeer Counter Gambit) 3. ed e4 4. d3 Nf6 5. de Ne4 6. Nf3 Bc5 7. Qe2 and then he played 7... o-o which I was sure was bad.

I was thinking I could just take his knight with my queen and then in order to get it back he would have to weaken the a2-g8 diagonal, which would be really bad for him. Play continued 8. Qxe4 Re8 9. Ne5 f6 10. Bd3 g6

I couldn't see the correct continuation here. I went and looked up the variation in "Nunn's Chess Openings" and it said that Estrin (former world correspondence chess champion) said White has a clear advantage after 11. Qc4. But I couldn't really see that either. Over a few weekends I let the chess program Crafty analyse the position using computers at work. It suggested the best sequence of moves after that is 11... Bd6 12. Kd1 fe 13. fe Bxe5 14. Re1 Nd7 15. d6+ Kg7 16. dxc7 Qf6. It's not clear to me that White's advantage is obvious in that position.

Recently my officemate has gone on holidays, and a PhD student we had has left, so there are computers sitting around doing nothing. I have been thinking about what I could possibly do with them on the weekend. I used to think that Crafty was one of the best programs around. It seems that's not the case anymore. There's a confusing array of rating lists and tournaments right now. It really reminds me of compression software. Every few months, a new program appears that just blows all the old programs away.

The two historic lists are SSDF and the Selective Search list. There's a program called Rybka by an International Master Vasik Rajlich which tops both of these lists, along with the SCCT, CCRL, CSS, WBEC, and CEGT rating lists. I'd have to pay 59 euro for the latest version though, and the free version (1.0 beta) is stronger than any other free program anyway, according to the CEGT list. I remember back in 2002 a program called Ruffian was released by an unknown author and became the strongest program around. Some established chess programmers were very skeptical that this was possible because experience had shown that previous programs which proved to be very strong were almost always clones or rip-offs of other programs. Strangely, Ruffian 2 is said to be no longer available.

The very latest version of Crafty, 20.14, is performing very badly in fast tournaments. In the latest World Computer Chess Championship and CCT Crafty's performance was only average, though it has won twice in the past, CCT-1 and CCT-6.

There are a confusing number of other programs to choose from -- there are even many open source programs stronger than Crafty now. I just looked at all the lists and picked the free ones which are above Crafty. That's Rybka 1.0 beta, Toga II, Spike, Glaurung, Pro Deo, Naum, Ruffian etc etc. I don't have a lot of time to think about which is best, so perhaps I should just stick with Rybka.

From July to October 1999, when I was a PhD student, I worked pretty much full time analyzing the Kasparov vs the World game. I used Crafty exclusively and often used the maths department's computer network to examine lots of positions simultaneously. But the computers I used back then are just a joke now, even compared to the computers in my office at work. "The World" was assisted by four young analysts and voted on which move to choose at each stage. One of the analysts, Irina Krush, worked very hard on the game with the assistance of her trainers and participated heavily in discussions on the bulletin board provided. If the World had followed her recommendations at every step, the game would have been a draw. I suppose the most interesting positions in the game are those where the four analysts recommended four different moves - and this only happened twice, at moves 15 and 16. I plan to analyse these two positions extensively over weekends using my wonderful work computer. Maybe I'll throw in another 2Gb of RAM so I can max out the RAM and have more hash, making the programs stronger.

There are many other interesting positions in the game and it is hard to rank them. As I followed the game and bulletin boards pretty much full time, I remember lots of places where obscure moves were suggested. Apart from the four official analysts, there was also correspondence grandmaster Roberto Alvarez, OTB GM Duncan Suttles, and "IM2429" suggesting moves that no-one else considered. I also have Kasparov's book on the game, which is unfortunately out of print now. I'll just list them here.

Move 6. The Russian Grandmaster Chess School strongly recommended 6... Nd4 and wrote that it "equalises the game completely". The official analysts recommended Ne5, Nf6 and g6.

Move 7. Again, the official analysts diverged and recommended Ne5, Nf6 and g6.

Move 10. The most exciting move -- Qe6 and o-o were recommended. o-o would have led to a boring game.

Move 15. Apart from the four analysts' suggested moves, Motylev and Kalygin have played ...e6, Werner has played ... Ne4 (originally recommended by the Barnet Chess club), and ...Ke8 looks reasonable.

Move 16. Again ...Ke8 looks reasonable. (and ...e6 too?)

Move 18. Krush recommended ...f5 and the other three analysts recommended ...Nd4. In his book Kasparov suggests ... Bd4 draws easily and ...e6 is solid.

Move 19. Krush recommended ...Qb4 and the other three analysts recommended ...Qd4 and ...Nd4. IM2429 and the Computer Chess Team recommended Be5.

Move 21. Felecan recommended ...Rh8 and the others ...Rxa4. Kasparov said ...f4 leads to a balanced position.

Move 25. Pahtz recommended ... d5 and the others ... Bd4. Kasparov wrote that ... d5 and ... Nd4 would draw too. Alvarez recommended ...b5.

Move 26. Krush recommended ... f4 and the others ... Bc5. Here Duncan Suttles had an unusual recommendation, 26 ...d5. And lately Crafty has been suggesting ... Ne5.

Move 29. Kasparov wrote that ... Qe2 is a draw.

Move 33. An interesting question is whether ...Bxg3 might draw, but perhaps computers will not be able to answer that question.

Move 34. Kasparov wrote 34...Bh8! is an easy draw.

Move 35. Kasparov wrote 35...Ne5 is a draw.

Move 36. Suttles recommended ...Bc3.

There were no more choices until move 51... at which point the World played the wrong moves and went downhill. Unfortunately I don't remember what Suttles, IM2429, Computer Chess Team, GM School etc were recommending at some points. I'll have to pull out my old files!

Back to the Falkbeer counter gambit position above. I've briefly run it through Toga II which seems to prefer 11. Nc3 Nd7 and thinks White is about a pawn up. Interesting.


Yazd from the Hosseinieh

A great place and my favourite city in Iran, but I wish I had worn newer jeans!