The BBP Algorithm
(or should I say, the Plouffe algorithm?)
I found a letter from Mathematician Simon Plouffe, discoverer of an algorithm to find digits of pi; it truly is a sad story...let it serve as a warning for everyone, especially aspiring technocrats
Originally posted here
The story began many years ago in 1974 when I wanted to find a formula for
the n'th digit of Pi. I was studying rational and irrational numbers. With
my calculator I was computing inverses of primes and could easily find a way
to compute those inverses in base 10 to many digits using congruences and
rapid exponentiation. Since it appeared impossible to do the same for Pi, I
wanted then to find a simple formula f(n) that could compute the n'th digit
of Pi. I had that idea for 20 years.
Since the computation of Pi looks more complicated than the number E , i.e.
exp(1), I studied a way to compute that number instead. At that time (around
1983), I had a simple Basic program that used a spigot algorithm to compute
E, as expected that algorithm worked but of course but was taking an
increasing amount of memory. My question was : why can't we do it for E or
Pi or any irrational numbers like sqrt(2).
It was during the year 1994 that I began to compute arctan series but I did
not realized that this meant a lot. I was able to use an algorithm to
compute arctan of 1/5 with fast exponentiation without realizing that it
could compute arctan(1/5) in base 5 very fast since the rapid exponentiation
was natural in that base.
Later in 1995, around august 7 of that year I suddenly realized that log(2)
was fast computable in base 2. Since I had a bit of experience with spigot
algorithms and also my little Basic program to compute arctan, it was not
difficult to adapt the algorithm to log(2). In the next few days I made my
first program : A program to compute log(9/10) in base 10 using a very small
amount of memory and very fast. The program had 432 characters long.
That discovery was a shock to me. I realized that I had found it yes but it
was not new to me since I could do arctan(1/5) easily too but it took me 2
years to realize it.
This is where I began to use Pari-Gp, that program could find an integer
relation among real numbers (up to a certain precision), very fast.
During my stay at Bordeaux University in 1992-1993 I perfected that program
I had that could interface Pari-Gp and Maple. That little Unix script had an
enormous advantage of flexibility because I could set up a series of real
numbers to test among 1 unknown. At that time I was beginning to find new
results, the programs were able to find identities.
That program was the one that found the formula for Pi in hexadecimal (or
binary). I also used another one : PSLQ. It was a good program but a bit
cumbersome to use since it is written in Fortran. Nevertheless I made an
interface to Maple too. Pari-Gp was by far easier to use and faster for
small cases (up to 10 real numbers at the time with 100 digits precision was
enough for those kind of problems).
This is where I made the biggest mistake in my life : To accept the
collaboration of Peter Borwein and David H. Bailey as co-founders of that
algorithm and formula when they have found nothing at all. David Bailey was
not even close to me when I found the formula. He was added to the group 2
months after the discovery.
I was naively thinking that I could negotiate a job as professor at Simon
Fraser University, which failed. I am very poor at negotiations. I remember
that day when the Globe & Mail newspaper article went out in October 1995. I
was at Jon Borwein's house and he had a copy of the newspaper in hand. This
is where I asked him to become a professor at SFU. He simply replied right
away < don't even think about it >. I thought, this is the best chance I
will ever have to become a professor there, since it failed, I decided that
I had to leave that place.
I was very frustrated at that time, in late 1995 after the discovery. I
realized that many small details where terribly wrong. They were getting a
lot of credit for the discovery and I had the impression of not getting
anything in return. My strategy failed. One of those details was the article
of the Globe and Mail, I asked Peter Borwein : why did they putted the photo
of you and your brother on the article ? Your brother has nothing to do with
this!. He simply replied that the Public Relations at the University made a
mistake. Later that year, I was invited to a ceremony in Vancouver for the
CUFA (faculty of the year Award). This is a prize with plaque and mention
that those 2 brothers received for the discovery of the formula. They simply
mentioned my name at the ceremony and I received nothing at all. They made a
toast to the queen of England, I did not stand up.
In late 1995, there was that Canadian Math Soc. congress in Vancouver, I was
not invited to talk about the discovery. There was even a guy (Stan Wagon)
that said to me, I don't know if you have anything to do with this but in
all case, this is good for you isn't ?
Then in 1996, I realized that if I get up at night to hate them it is a very
bad sign, it means that I have to leave that place (Simon Fraser
university). I was convinced I had no future at all with those 2 guys
around. I was making serious plans to leave.
The story of the formula (my formula), was not the only one. The same thing
happened with the ISC (the Inverse Symbolic Calculator). The story is even
more ridiculous. I opened the site with my constants in July 1995 and it was
an immediate success. The 2 Borweins had nothing to do with that thing, I
had made the tables and all of the Unix programs to run it. The precious
help I had was from Adam Van Tuyl, a graduate student, he made most of the
code behind the web pages, later Paul Irvine made some additional code.
At that time the local administrator of the lab. tried to convince me to
stay even to pay me for maintaining the ISC, I refused. I wanted to leave
with what I had : my tables of real numbers and sequences I worked for years
(since 1986). This is why I opened the Plouffe Inverter with my name in
1998, to keep what was mine. When I realized that I was about to loose the
paternity of the ISC, I left in march 1997. I went to Champaign Illinois to
work for Wolfram and Mathematica. (this time it took me less time), that one
was worst than the 2 brothers combined. I simply left as soon as I could, 5
months later.
Peter Borwein wanted very much that I do a Ph. D. on the ISC but he wanted
also to publish (with his name of course) an article before I deposit the
thesis. Again the same story was going on, these 2 guys are so greedy I
can't believe it. The behavior they had with me was not exclusive,
especially Peter Borwein he was the same with most of his students,
especially the good ones, sucking the maximum. Jon is the same but he has
more talent in politics (more money too). He is good but has a tendency to
site himself a lot. He thinks that if he had the idea of the sum of 2
numbers at one point in his life then all formulas in mathematics are his
own discovery.
About David H. Bailey. He came after the discovery of the formula and my
small basic program , I had also a Fortran version. This is where Peter
Borwein suggested to add him as a collaborator to the discovery since he
contributed to it (as he said), this is my second big mistake. Of course he
accepted to co-write the article, who wouldn't ?! David H. Bailey (and
Ferguson) are the authors of the PSLQ program. That program is the
but what made the discovery was Pari-Gp and Maple interface program I had.
So actually, that person has nothing to do with the discovery of that
algorithm and very little to do with the finding of the formula. The mistake
was mine. Saying that Bailey found the formula is like saying that the
formula was found by the Maple and Basic program.
I tried very hard to correct the situation avoiding the subject of the
actual discovery of the algorithm and the formula, I made an article in 1996
for the base 10. I thought naively again that this would re-establish the
situation, it did not. I almost accepted to do a film at one point in 1999
when a certain guy from England that wanted to make a movie on Pi and the
discovery of the formula. he asked me if I would accept to talk about my
should had. There was that book of Jean-Paul Delahaye (le fascinant nombre
pi) that mentioned the Plouffe algorithm and formula because I told him part
of the story. In some way I was afraid of revealing that enormous story.
Why was I so naive ? I had a previous collaboration with Neil Sloane and the
Encyclopedia of Integer Sequences and the web site, this was really a big
success and Neil is the person I respect the most in mathematics so this is
why I thought (wrongly ) that my collaboration with the Borweins had to go
well, a big mistake.
Why do I write this ? To tell the truth and also the arrogance of those
people makes me sick.
Will I gain something from this ? I don't care, I have nothing to loose.
Simon Plouffe Montréal, le 22 juin 2003.
####################################################
No comments:
Post a Comment