Sun Mon Tue Wed Thu Fri Sat
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
The world of carton
When Annabel heard about the
design contest The world of carton (De wereld van karton in Dutch), she quite quickly
got the idea to transform a drinking package into a
general purpose holder that can be used to hold a
mobile device close to the wall when it is being charged.
It is a simple but clever design, I think. She joined
the contest, but might be just too late, because the
closing date is today. On the right top a picture of
her design in case it does not win a price. The picture
on the right shows her design.
Today, we had our yearly Skûtjesilen trip. (Except for
last year, when we went to Barcelona.)
We left around 8 o'clock from the office and arrived in
Elahuizen around 10 o'clock. We sailed with three skûtjes.
I went with the Eelkje II again. We sailed from 11 till
1. I stayed on the front of the ship helping with the the jib
(foresail). Then we had a packaged lunch at an island. We sailed
through some canals and had to use the engine for some part.
We also had an extra stop for some icecream, a reward for a
conteste that some colleagues won. When we left around four,
the wind was quite strong, and we did some real sailing. I
stayed at the back of the ship with the captain and helped
a little. Then around half past four, we noticed a skûtje
that had capsized. (See the picture on the right.) It was quickly
surrounded by some other skûtjes. First we thought that it
was not one of our ships, but after a phone call it turned out
that it was one of ours. We lowered the sailed and went back
to collect our colleagues, who had already transfered to another
skûtje. Not all of them had fallen in the water. Luckily,
there were no personal injuries, except for some small cuts.
But almost all of them had wet feets. Luckily, we had some dry
sailing clothes on our ship. The water was not that cold, but
staying in the wind with wet clothes is not good. Later all
the personal belongings where collected from the ship. Some
electronic devices did not survive the water and died. We had
some BBQ. And a close colleague of mine, who deals with the
helpdesk, was nominated for employee of the year. Then we heard
that the skûtje had sunken shortly after it had been
erected again, because water flowed into it. The lake is at
most 3 meters (10 feet) deep, and they where going to try
to salvage it with some crane ships later that evening.
Around 8 we left for home. We arrived in Enschede around 10
o'clock. (The KML file of todays
GPS tracks to be viewed in Google
Earth or shown with Google Maps.)
Two support beams
A few minutes after noon, we heard a loud rumble outside.
Yesterday, a colleague of mine and I had watched the cranes beside
De Grolsch Veste, the stadium of the
soccer club F.C. Twente, slowly swinging due to some strong winds, and
discussed the possibility of them crashing into the stadium. It
immediately came to my mind up. I jumped to the window and saw the
last second of some part of the roof collapsing. Everybody looked
out of the window. I took two pictures. For our point of few
we could only see part of the roof, as our view is obstructed
by another building. I wrote some messages on twitter including
one (in Dutch) mentioning that two support beams
(twee steunbalken in Dutch, had collapsed. I meant to
say: two roof beams. Shortly after that, I discovered
that the whole roof on one side (except for the corner) that
was being constructed, had collapsed. I wrote another tweet
to correct my previous tweet. A little
later, I discovered that the expression twee steunbalken
had been copied in a very brief news item of Twentse Courant Tubantia, a local newspaper. I strongly
suspect that this was taken from my tweet, and I wrote a comment
to remark that this was wrong. Comments are always reviewed
before they are shown, so, I had expected that the reviewer
would have noticed my remark and correct the mistake, but instead
the comment was placed below the news item. During the afternoon,
I noticed that the expression twee steunbalken was copied
by many other news websites. I even found something resembling
this on Wikipedia and removed it. Later on the day, I also found
the expression two support beams on many websites around
the world. I am afraid that all those uses can be traced back
to me using the expression twee steunbalken on twitter,
as I was the first to use it.
Horizontal and cross beams
It is becoming clear that the collapse of part of the roof of De Grolsch Veste is due to missing horizontal and cross beams. Possibly
adding the additional weight of a large screen on the edge of the roof might
also contributed to the collapse of the roof. Although it should be remarked
that this screen already had been there for some weeks. It seems that seven
very solid horizontal beams connecting the most outside points of the
roof trusses where still missing. Also two horizontal beams on the top of
the roof trusses where missing. From an important combination of four
cross beams, three beams where still lacking. Of another fourteen cross
beams only four where placed so far. The theory that the missing beams
where essential for the collapse to happen is supported by the fact that
the first roof truss that did not collapse, had horizontal and cross beams
on one side attached. From some airial pictures it is clear that some strong forces worked on
this truss, but nevertheless it looks unharmed.
Another interesting picture shows the collapsed roof construction
from the outside. Seven roof trusses where damaged. With four the most
outside node collapsed to the right, The following two collapsed to the
left and the last did not collapse on the outside node. Which is rather
surprising. It could be explained by the fact that the horizontal beam
of this truss is disconnected at the spot where it should have been
connected with bolts. It would be interesting to know when this connection
snapped, at the beginning or at the end of the collapse. What surprises me
about the picture is that how little the trusses moved to the side. From
the fact that they only had to move so little to the side, (I think) one
could conclude something about the forces on the trusses, especially for
the sideway force needed to make the trusses collapse. It is possible
that only a slide horizontal movement (of the edge of the roof) was
sufficient to make it collapse due to the instability of the roof
construction. I think it will be interesting to know the construction
planning and to find out why the horizontal and cross beams where not
installed yet, while the construction was put under load. Who was
responsable for the planning and/or was there any deviation from the
In the evening, Annabel and I,
cut down our oak tree in the back garden.
It grew from an acorn that Annabel put in the ground six or
seven years ago. Now it was about 5 meter (15 feet) tall. The
problem is that it is very close to the wall of the extention of
our neighbours, and I am afraid that it will damage the foundation
in due time. (Picture from
June 6, 2010.) I also cut some of the roots going out from the main
root. Our neigbour also helped a little to cut the trunk in some
smaller pieces. Annabel found a very small oak tree on the lawn.
I don't know how it got there, but I will cherish it.
Katla showing activity
It seems Katla is showing
some activity lately with swarms of earthquakes and harmonic
tremors. I found a good blog: Iceland Volcano and Earthquake blog.
This morning, I accidently found the PhD thesis Monte-Carlo Tree Search by Guillaume Chaslot in some bookshelve. I wondered how it got
there, as I could not imagine that anybody in the office knew
what it was about. Then I heard that a colleague of mine often
received PhD thesis because his involvement with some kind of
board. I asked him of I could get it and he replied: Sure!
The subject of the thesis is related to computer programs
playing Go. I am also interested into
the subject, because it might be a technique applicable to
Havannah as well. There seems already to be some work on this:
At home, I started reading the thesis.
Computer learns language by playing games
Today, I came across Computer learns language by playing games, which links to the
paper Learning to Win by Reading Manuals in a Monte-Carlo Framework.
I submitted a story to Slashdot,
which got accepted with some slight editting.
I received an email from Christian
Freeling in which he states that the abstract game Symple will be the next major challange for the
AI community. A decade ago, he predicted that the abstract game Havannah would pose great problems for the AI
community. That indeed proved to be the case and it still holds, but,
as he acknowledges, Monte-Carlo Tree Search and UCT (Upper-Confidence-Tree)
have given the efforts a firm handle. Already good results have been shown
for computer programs playing Go. Havannah received
less attention so far, but maybe that may change when his 1000 Euro challenge
is coming to a close next year. Symple indeed poses a new challenge because
of its extreme branching factor in the middle phase of the game, when there
are in the order of 1020 different moves to consider. The reason
that there are so many moves, is because each move can consist of placing more
than one stone depending on the number of groups you have. With Symple
you can either decide to create a new group or to (optionally) grow each
group on one of its direct neigbours. But maybe it is possible to use a
Monte-Carlo technique to determine which is the best move for each group and
from this information construct a best multi-stone move, or at least limit
the search space by trying various combinations of good moves from each
Today, I visited bookshop De Slegte rondgekeken.
Outside on a table I found Een zomer lang, the Dutch translation of
by Truman Capote,
ISBN:9789023420590. At 11:57 bought this book for € 3.49.
MCTS and Symple
I finished reading the thesis Monte-Carlo Tree
Search by Guillaume Chaslot, ISBN:9789085590996. I did not read every letter,
because the thesis does contain a lot of repetition, such as the final
chapter, which repeats all the conclusions that where already given at
the end of each previous chapter. I also did not understand every part
of it in great detail. I have to say that it is well written and that it
has given me a clear understanding of what Monte-Carlo Tree Search (MCTS)
is and how it is applied in programs playing Go.
What I have learned is that although MCTS is a great technique it still
requires a lot of domain language including expert knowledge from the
game in order to create a stong game playing program. I am afraid that
the results that have been achieved with Go, cannot be directly applied
to a program playing Havannah and that
Symple will be a great challenge.
After I wrote my previous entry about Symple, there
has been some discussion with Christian Freeling and Benedikt Rosenau.
I realized that my idea would work good in the later phases of the game,
but that it would likely fail in the first half of the game. And it should
be noted that it is this half which is the most important, because Symple
lacks any 'dramatic' aspects, such as capturing. Also there is not much
room to fix 'errors' made earlier on in the game. This makes it a very
Havannah and Fuego
Last year, the CMPUT 655 - winter 2010 course from the Department of Computer Science of University of Alberta
had a Havannah playing program as an
assignment. A sample playing program based on Fuego was given.
Fuego is a framework for
Monte-Carlo Tree Search playing progams. The course was given by
Müller, who is also involved in the development of the Fuego
Birthday party Meindert
Today, Andy and I went to the birthday
party of my old friend Meindert.
There we met with his family, some of whom I had not seen in more
than 20 years, and the family of his girl friend and some other
friends. Andy did behave well. I played some connection games
against Ton. In the past we often played Go.
He now is really into connection games, such as Hex, Havannah, and Atoll.
He is selling boards for
those games. He is also one of the best, if not the best,
Havannah player in the world. I lost from him with three stones
ahead. I also asked him about the Havannah playing programs, and
he thinks there is a good chance that Christian will lose his bet.
Because of a road being blocked we took a very different
route home. I decided to take the highway instead of following
an obscure detour like we did on our way going there.
KML file of our trip which can
be viewed in Google Earth or
directly in Google Maps.
Extremely Loud & Incredibly Close
Yesterday, I finished reading
Extremely Loud & Incredibly Close by Johanthan Safran
Foer, ISBN:9780141025186, in which I started reading on April 10. On
Thursday, December 23, 2010, I bought a
second-hand copy of the book from bookshop
De Slegte for € 4,50. On of the reasons was that I was
intrigued by typographic experiments, such as the use of photographs,
pages with a single line, and alternative spacing. The book even
contains a short movie on the last 14 pages of the book. Yesterday and
today, I tried to decypher the phone spell digits given on pages 269-271.
The grandfather of the main character, who cannot speak, is trying to
get a message across by typing the digits matching the letter after he
has arrived from abroad on the airport. Last week I already started
working on a program to decode
the digits using a word list. I downloaded the SCOWL from Kevin's Word List Page and used some parts of it. Yesterday,
I scanned the three pages containing the numbers. The I used FreeOCR version 3.0 to get some raw OCR output, which was not of a very good quality. But what
can you expect from a free package. I spend some hours correcting
the output using the MySample editor, which
can display both images and text. Matching-up 'lines' from the scan
with the text was relatively easy using Ctrl-Tab for switching between
windows inside MySample and using the arrow keys to move around the
image. Next I wrote a small program
to transformed the corrected OCR output
to a file containing the 'sentences' assuming that the question and
exclamation marks are used for termination of sentences, or at least
occure on word boundaries. When studying the sorted output of the program, I found one or two errors in
the corrected OCR output. But it also seems that some errors where
introduced when type-setting the book. There seems to be one '3' digit
missing between pages 269 and 270. Somewhere else, an '3' seems to
be accidently changed into a '7'. I also found How we work: Jonathan Safran Foer, author, which contains
some remarks about the digit sequences stating that only the first
sentences have some meaning, and that the rest is just random sequences
being copied. You can have a look yourself by studying the results yourself.
Today, I went into the city with Andy.
Of course, I also visited some bookshops. First we went to bookshop Kruimeltje, where I bought the PhD thesis:
PARELLA: Measurement of Latent Traits by Proximity Items by
Herbert Hoijtink for one Euro. I also walked around the second floor
with Andy, which is kind of interesting, because in a way it is like
a gallery along the walls with even a small 'bridge'. They have a
nice collection of rare books by Harry Mulisch.
We also went to bookshop De Slegte.
Andy looked in the childrens department and later on the second floor
he looked through their collection of second hand comics.
At 12:33, I bought the following two books:
This months interesting links
| June 2011
| August 2011