Sun Mon Tue Wed Thu Fri Sat
1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31
|
Castro playing Havannah
This morning, I received an email from Timo Ewalds, the author of Castro,
a Havannah playing program, which seems
to be strongest at the moment. (Timo is the founder of
Nexopia.
His profile on LinkedIn).
His program was the strongest of the six Havannah playing programs developed during
the CMPUT 655 - winter 2010 course from the Department of Computer Science of University of Alberta (which I mentioned on
July 18). After this he continued developing his
program further. Currently, it runs as a bot on Little Golem. Havannah playing bots on Little Golem with there descriptions:
- Castro,
where it is described as: Havannah program written by Timo Ewalds at the
University of Alberta, Canada. UCT based, with rave, some knowledge, and a few
other tricks, written in C++. Currently accepting unrated games on all sizes,
and rated games on the smaller sizes for now. Used 90s per move for the first
few games, using 5 minutes per move now.
- wanderer_c:
An Havannah bot written by Richard Lorentz
with the help of students. Thanks to klaashaas for providing the Ruby s-cripts that
allow wanderer_c to work without any human intervention and thanks to Richard Pijl
for his technical advice.
- Deep Fork:
Havannah AI by Thomas Reinhardt, HTWK-Leipzig, Germany
(Source: Havannah 2012 human
vs. bot challenge.) See also this discussion about Castro.
Timo also pointed me at the master thesis Knowledge-Based Monte-Carlo Tree Search in Havannah by
J.A. Stankiewicz, which was done at the Games and AI Group of the University Maastricht. It is possible that
another strong Havannah playing program may result from this.
Repeating sequences
In the past week, I spend a lot of time (probably too much) in developting
a program for analyzing the phone spell digits
sequence in the novel Extremely Loud
& Incredibly Close. The program underwent many changes while
discovering structures in the sequence. It became clear that the sequence
(spanning about two and half page) was largely created by making copies,
trimming those copies and making some (possibly involuntary) mutations.
I worked towards a program describing the copying (with changes) as clearly
as possible in written language. This took a lot of time. The final program
can be found here and the results
here. Now I still have to interpret
these results and make an attempt to reconstruct the order in which the
copies and mutations where made.
Around noon, I finished reading Called Out of
Darkness: a spiritual confession by Anna Rice, ISBN:9780307268273. Last Sunday, I started reading the book
after I bought it last Saturday for € 8.95
from bookshop De Slegte. The book is
an autobiography with respect to her faith. The first half she deals with
being an orthodox Roman Catholic in her first twenty years of her life.
She became famous for writing vampire fantasy and being an atheist. But than
the book describes how she became a christian again and rejoined the Roman Catholic
church and went to write some novels about the life of Jesus based on the
new testament gospels. Her conversion to christianity seems to be based on an
deep inner conviction and not so much on rational arguments. Her faith appears
very genuine, and somehow, I am not surprised by the fact that on July 29, 2010,
she publicly renounced her dedication to her Roman Catholic faith (which is
not mentioned in the book), because she is one of those rare independent
thinkers who arrive at their convictions all on their own. An interesting
book to read.
Books
This afternoon, at 12:34:50, I bought the book Oeroeg by
Hella S. Haasse, ISBN:9789059651012, from the books department from de Bijenkort for five Euro (including
50% discount).
I also got three Star Trek novels (from The Original Series) which
she had bought in the United States. These are:
- The Vulcan Academy Murders (#20) by Jean Lorrah. ISBN:067164744X.
- The Final Nexus (#43) by Gene DeWeese. ISBN:0671660187.
- Rules of Engagement (#48) by Peter Morwood. ISBN:0671661299.
More Havannah bots
Today, I got an email from Marcin Ciura informing me about two othere Havannah playing bots on LittleGolem. These are, with their descriptions:
- Ring von Fehler: Havannah robot programmed by Johannes Waldmann at HTWK Leipzig,
Germany. Key features:
- UCT based
- heavy playouts (checking for forced moves) but these are slow, so the tree is small
- static evaluation (comparing lengths of shortest paths on board)
- implementation uses C for low level stuff (playouts, path lengths) and
Haskell for everything else (tree,
HGF parser, LG connection)
Will happily accept all challenges (but probably play miserably on larger boards).
Thinking time per move: boardsize minutes (that is, 4 minutes for size 4, etc.) Actual
response time: depends on number of active games. Expect frequent irregular behaviour
and breakage.
- Lajkonik:
A Havannah bot programmed by Marcin Ciura and Piotr Wieczorek from Kraków.
Using MCTS with transposition table and RAVE; written in C++. Currently
runs single-threaded on a 2.53 GHz Intel Core 2 Duo, achieving 60000 light
playouts per second on an empty board. An earlier version placed third out
of five programs in the Havannah tournament at the 2010 Computer Olympiad in Kanazawa.
This instance only plays on size 5 board. It thinks over each move for five
minutes. No more than one process runs at a time: games in progress with
Lajkonik's turn to move are queued.
The current ranking of the five Havannah playing bots at LittleGolem is:
Demolishion roof Grolsch Veste
In the past two weeks, they worked hard to demolish the collapsed
part of the roof of the Grolsch Veste
as can be seen in the picture on the right. What surprised me that
they also removed the corner of the new extention which did not
have visible signs of being damaged. I guess they wanted to be on
the safe side.
Zomergo
Today, I went to ZomerGo, a five day Go
tournament in a relaxed atmosphere. Only one round is played per
day with a playing time of one hour. We are staying in
De Windwal
in Lunteren. I arrived five minutes before the registration would
end, but everybody was still waiting outside and registration
started only about half an hour later. The first round was only
played around four o'clock in the afternoon. I played against
Michael and already got behind very soon, and lost the game, but
I did enjoy the playing very much. It is the first time that I
played since WinterGo last
December. Everything was delayed, and dinner was served around
nine o'clock in the evening.
ZomerGo - Day 2
This morning, I played against Francien in the Go tournament. She got some control over the center of the
board, but I managed to invade it and ended up winning the game.
We played without clock and finished our game well within the
theoretical maximum of two hours. Afterwards we played another
game, which she won. Than we together played against Thomas, but
we lost with many points, more than one would expect from our
playing strength. It seems that cooperations did not really made
us stronger, maybe because we were focussing too much on tactical
moves.
Copy analysis
This morning, I finished the copy analysis of the the
phone spell digits sequence in
the novel Extremely Loud &
Incredibly Close. Yesterday, I found a small mistake
in the corrected OCR. It seems
that the author has created the two and half pages of phone spell
digits by making repeated copies of earlier parts and introducing
some errors (on purpose) especially for the first part and inserting
some words, in particular the digits for the word "love". However,
there are some exceptions, as sentences 34 to 39 seems to be copied
from sentences 44 to 49. Also there are two overlapping regions from
which the sentences 93 to 103 could copied from. This could be purely
accidential, or it could be that the sentences 90 to 117 where copied
from an earlier part where the author later removed some scenteces
from (at the start or the end).
Below the results are presented. Each sentents is identified with
number prefixed with 'S'. The period notation is used to identify
a specific character in the sentense. Ranges within a scentense are
placed within square brackets. For all ranges the last scentense or
character is to be included.
- S1 is '696263473543258626345878277483328843247767846333863463673465357!'.
- S2 is '6432267425638726343?'.
- S3 is '5763' and S1[13-48] with 1 mutations.
- S4 is S3[25-34].
- S5-S7 is copy of S1.45-S3.24 with 3 mutations.
- S8 is S1[25-58] with 4 mutations and S1[25-58] with 7 mutations.
- S9 is '526265952?'.
- S10 is first 27 of S1 with 8 mutations and '42552924526'.
- S11 is '42' and S10[3-36] with 5 mutations.
- S12 is S11, '472272', and S8[52-66] with 2 mutations.
- S13-S21 is copy of S4-S12 with 15 mutations, "love?" inserted before S7, and
S8.30-S10.23 replaced by '!'.
- S22 is '65557!' (or end of S5 with 1 mutation).
- S23 is first 11 of S6 with one mutation and '526'.
- S24 is S20[15-19] with '?'.
- S25 is '5765', end of S20 with 1 mutation (S20 = S11), '455652', and S21[21-27].
- S26 is S25[25-32].
- S27 is "love?".
- S28 is S3 with 9 mutations (or S13 with 3 mutations and start S14).
- S29 is S28[25-36].
- S30 is "love?".
- S31 is "love?".
- S32 is "love!".
- S33 is first 28 of S20 with 2 mutations, '4524', and last 22 of S7.
- S34-S39.5 is copy of S22-S25.29 where "!love!" is inserted before S25.30.
(Or S34-S39.5 is copy of S44-S49.7 where two '?' are changed into '!'.)
- S39.5-S43 is copy of S8.27-S12.
- S44-50 is copy of S22-S26 where "!love?" is inserted before S25.30.
- S51-S52 is copy of S10.28-S11.7 (or S19.6-S20.7) terminated with '!'.
- S53 is copy of S28.
- S54-58 is copy of S13-S17 where last '!' is replaced by '?'.
- S59-S66.29 is copy of S21-S28.29.
- S66.30-S72.4 is copy of S49.5-S55.4 where S51-52 replaced by '6545!45?'.
- S72.5-S78 is copy of S6.6-S12.
- S79-S85 is copy of S44-S50.
- S86-S90.5 is copy of S14.5-S19.5 where '!' at the end of S14 is replaced by ','.
- S90.6-S103 is copy of S72.5-S85.
- S93-S117 is copy of S40-64.
- S118 is "love!", like S65 or S27 with '!' instead of '?'.
- S119-S124 is a copy of S28-S33.
- S125-S129 is a copy of S92-S97 with S94.8-S95.24 replaced by '5652465552'
(which is the end of S116, S102, S84, S63, S49, or S25).
(Corrected copy analysis.)
This morning, I played against Hans in the Go
tournament. He recorded the game (SGF file). We
played very quietly for almost two hours. At the end Hans won with four
and half points. I could have killed one of his corners.
In the evening, the game Atlantis was played in the main room with six people on two
large tables put together. I watched the game being played a little. A Dan
Go player remarked that he thought it was the most briliant game after Go.
But then a 5 Dan Go player played against the latest version of the Go program Steenvreter (by Erik van der Werf with a five stone handicap. The program lost and at the end
started to play very desparate moves. Next, I went back to the Atlantis
game. After about half an hour, they decided to quit the game, because it
was quite obvious who was going to win the game anyway.
The Vulcan Academy Murders
Early this morning, I finished reading
Star Trek: The Vulcan Academy Murders by Jean Lorrah, ISBN:067164744X,
in which I started reading on August 6, the day I got
it from Annabel (together with two other Star Trek books, which she
bought in the United States for one dollar). I started reading the
book, because I wanted to know the quality. To my surprise it was
beter than I had expected. Of course, it is a classical 'Who did it?'
story. But there are some interesting side stories, which makes it
an interesting book.
This morning, I played against Peter in the Go
tournament. At the start of the game there where no confrontations,
and because I did not feel very happy about this, I decided to
cause some excitement by an invasion. But this turned out into
a big fight in the center, during which I was able to capture some
stones. I was not really at peace and at one point felt my heart
pounding. Maybe it has to do with the fact that I stayed up late
last night, but I could not really find my balance. I also lost
one group in the corner. At the end, I lost with about 45 points,
while Peter had enlisted himself as two degrees weaker than I.
I am getting the feeling that I am no longer a 10 kyu player.
Or maybe he has become stronger because of paying attention during
all workshops and discussions with stronger players. Afterwards
I walked around the woods with someone else and that helped clear
my mind.
In the afternoon, some games were discussed outside. I watched
the start, but started to doze away, and went inside. Later when
I returned, I noticed that the game I played yesterday against
Hans was being discussed.
First game of Atlantis
Yesterday evening, just before ten o'clock, they asked me to join
in a game of Atlantis because they still needed one player. I decided
to join. I mostly followed the example of the strongest player on
the table. But that was also the idea, so that nobody would accidently
getting killed by some stupid move. I tried to do some other things
while waiting for my turn, but that did not really work. Around half
past two this night we decided to quit the game, although we had not
finished it. Already one person had gone to bed, and we decided that
it would be rather clear who had won. I became third or fourth. Just
before the end of the game, I took the picture shown on the top.
This morning, I played against Philomeen in the Go
tournament. Because of the difference in playing strength, I had to
give her nine stones ahead an fourthy komi (points) ahead. Because she
is a lot younger to me, I decided to play easy on her. At some points,
I told her to be careful, but sometimes she still made the wrong move.
I killed one group of her in a corner. When we started counting, I was
surprised by the big difference. I did win easily, but it was the only
game that I would not have mind to have lost. She took her loss easily
and I hope that she did learn something from it.
Interim report Grolsch Veste
Yesterday, the The Dutch Safety
Board issued an interim report (in Dutch) about the collaps of the
roof of the new extension of the Grolsch Veste.
This report states (my translated from Dutch with help of Google Translate):
The basic construction of the roof was not completed yet before work began with
the finishing. Several essential elements for the stability (torque rods and
braces) were still missing: several braces in the roof, four braces in the
diagonal plane (wind bracing), all horizontal link bars on the back and
a torque rod in the roof between the trusses of the structure. In addition,
the anchoring of the trusses on the back of the stand were not fully
stabilized.
Before and during the collapse, construction was in full swing.
Roof sheeting was applied, the first missing link in the stabilization
diagonal roof was installed, under the roof work on the cabling and drainage
was done, and seats where installed under the construction.
Also the scoreboard had already been installed.
When assembling such structures, it is a given fact that the dividing line
between stability and instability is small until the last element is applied.
During the construction of the extension of the roof of the Grolsch Veste,
by a combination of the lack of essential elements for the stability and
a vertical load (over a long period) or an unexpected lateral load, the
turning point to instability was exceeded.
The report has an interesting picture taken from above the part of
the roof that collapsed, clearly showing the lateral movement of the edge
of the roof. The report also shows two diagrams about the missing torque rods
and braces. Red dotted lines indicate the missing parts. The finding presented
in this report are in agreement with what I wrote on Saturday, July 9.
A colleague of mine pointed me at a page of Olthof about constructing the first part of the roof.
The pictures on this page show how roof sheeting was applied before the
construction was completely finished. Possibly the horizontal link bars
where not installed so that the cranes could reach the farthest edge of
the roof. This shows that the same construction order was used as with
the earlier construction. But there is one difference, and that is that
now there was a large video screen attached to the edge of the roof.
There are some pictures taken half an hour before the collapse, which show that
a crane was positioned close to the screen, suggesting that it was just
being fitted to the roof, but that is not true. It already had been there
for some time.
Today, when at bookshop De Slegte, I found
that there was only one copy left of The Unseen University Cut Out Book by Terry Pratchett, Alan Batley, and
Bernard Pearson,
ISBN:9780385609449. I bought one copy on Saturday, July 30. But I did not want to destroy that copy for creating
the Unseen University. The last copy left was the viewing copy, and I only had
to pay af the price. At 18:58, I bought it for € 6.49. It a little
damaged, but that doesn't matter, because I intend to cut it to pieces
anyway.
Chestnut
Today, I went into the city, and while walking from bookshop De Slegte to
the day market, I crossed De Klokkenplas, a small sqaure in the center
of the city. There I found a chestnut, one of the very first, below a
large chestnut tree. Last year, on Wednesday,
September 22, I also picked a chestnut from this tree.
Rebuild collapsed roof
This afternoon, a colleague noticed that a new truss was placed on
the roof of the Grolsch Veste. This truss
replaces one of the trusses that was removed after a partial collapse on
Thursday, July 7, 2011 of the roof of
the new extention of the soccer stadium. In the past days, some parts
of the concrete (on which the collapsed roof fell) have been replaced.
This months interesting links
Home
| July 2011
| September 2011
Monday, August 8, 2011
Saturday, August 13, 2011
Sunday, August 14, 2011
ZomerGo - Atlantis
ZomerGo - Disaster
Wednesday, August 17, 2011
ZomerGo - Final day