Unbecoming Levity

Briaver and the Lynneaputians

Abacquer — Sat, 26 Jul 2008 21:29:08 -0400

So my friend Brian had an idea for a fun photo he wanted me to create for him. He wanted to be Gulliver of Gulliver's Travels, tied down on a beach by the Lilliputians... all of whom would be played by his wife Lynnea.

I originally wanted to shoot this on a beach (as per the original idea) but the lake in Chelmsford, MA that we were relaxing at didn't afford a beach with a suitable layout for the shot. So we instead did it on a nearby grassy knoll.

First we "bound" Bri by encircling him over and over with a thin twine. Then I helped him lie on a slightly raised knoll and lay flat before him so that the camera would be angled up to make him appear bigger. I said "imagine there is a six inch high person standing on your chest lecturing you" so that he would look in that direction. That was the first shot.

The subsequent shots all featured Lynnea using the thickest rope I could find on short notice as a prop. Bri would stand out of shot (typically on a footstool) and hold the rope while Lynnea would pull on it this way and that.

After recording about a dozen different poses, we did the lecturing pose, and then I shot the "sitting on the toes pose" when Lynnea was just relaxing on the stool. I thought the pose would come in useful and as you can see it did.

Shots were all done by daylight, no flash. EOS 5D with the EF 24-70mm 2.8L lens, which is the widest I own.

Then, after some cleanup in lightroom, came the hours of photoshop work to carefully clip Lynnea out of her surroundings in various pictures and edit her into this one at reduced scale. I thought the grass, which to a Lilliputian should be knee high at least, would be a problem, but it turned out that just using a gradient transparency on the ends of her legs (or whatever was closest to the ground) worked fine unless you look really close.

Things I would do differently if I shot this again:

I really need to get a chromakey backdrop for shoots like this. I shot Lynnea against a grassy green background, but that was not uniform enough to make clipping her out simple... it was fun, but it was a LOT of work.
Lynnea had been swimming prior to the shoot, and threw on a pair of pants for her poses. But in each subsequent pose water slowly seeped through the material and created spots in various places. I was mostly able to edit those out, but it was additional work.
Get thicker rope, or simply edit the rope out altogether and use the twine. The size difference between the reduced rope and the twine bugs me a little.
A beach location with a nice uniform ocean background would have made for easier editing.
I would have backed off a little more when shooting Bri. At 4x6, 6x9, or 8x12 the photo is fine, but at 8x10 the ends get clipped. That was just dumb on my part... to produce an unclipped 8x10 print some edits would be required.

All that said, I am really pleased with how the resulting image came out. And more importantly, my beloved friends Brian and Lynnea are happy with it, which is really what this was all about. Doing something nice for people I hold very dear in my heart.

Love you guys, glad you liked the photo!

Das Rad

Abacquer — Mon, 14 Apr 2008 19:46:54 -0400

Here's a funny animation I caught on Pharyngula, the excellent science blog by P.Z. Myers. The audio is German, but there are subtitles. I got a kick out of it, perhaps you will too?

Das Rad

Mark Isaak's Index

Abacquer — Wed, 16 May 2007 09:58:34 -0400

Wow, I just stumbled across The Index to Creationist Claims by Mark Isaak. What a great resource!

The Amazing Color Changing Card Trick

Abacquer — Sat, 05 May 2007 00:18:27 -0400

I found this awesome video via the equally awesome Bad Astronomy blog. I'm not big on card tricks, but this one is really amazing, and you might learn something from it as well.

Weelanders Show How Evolution Leads to Extinction

Abacquer — Sun, 18 Feb 2007 22:12:45 -0500

Okay this is going to be a lengthy article with a lot of graphics, so I am going to use an excerpt today. Basically it's a summary of the first run of my Genetic Factoring sim, and how it demonstrates that evolution can lead to extinction. It also includes a summary of the most efficient (and robust) genome to be generated by random mutation and natural selection, and how it compares to the original genome. Let's begin with a graph...

Genetic Factoring -- Weelanders Run Amok

Abacquer — Sun, 18 Feb 2007 10:00:00 -0500

I'm sure you remember the Weelanders... artificial life forms which lived on a grid and helped demonstrate that evolution through random mutation causes natural selection to kick in something fierce. At the time I originally ran tests with the Weelanders, I hypothesized that the Weelander concept could be adapted to do more than just demonstrate natural selection.

And, a few months ago, I did just that. I built a new version of the Genome application called "GeneticFactoring", and stripped out a ton of Weelander support code. I wanted to make Weelanders that factored numbers and then reproduced based on how successful they were at factoring. As a result they didn't need to search for and consume food, they didn't need sexual reproduction, they didn't need the ability to move, heck they didn't need chromosomes in pairs, as a result, all of the "critter emulation" code was unnecessary, resulting in a simpler Weelander support system.

What they did need was the ability to compute square roots, find primes, and factor numbers... this was the new code that needed to be added, and it was all going to be written in the native Weelander instruction set. It took me awhile to code the genome "Facto-f", but eventually I had it working. Then I modified the container and world code to invoke the Weelander factoring code in the following manner:

Populate: at the beginning of time, populate the grid so that it is completely full of the starting Weelander genome "f".
Setup: build a list of 50 numbers to factor, then factor each number using internal container code and store the results.
Execute:
1. for each critter on the grid
2. run through the 50 numbers
3. call the Weelander's Execute method and pass it the number to factor.
4. The Weelander factors the number and returns the results.
5. If the Weelander goes into an infinite loop without returning results, terminate it.
6. Otherwise, compare the Weelander's results to the original results.
7. If they do not match, terminate the Weelander for inaccuracy.
8. If they do match, note how many execution steps the Weelander needed to compute the result, and proceed to the next number.
9. Once all the numbers have been factored, compute an efficiency value for this critter (total number of steps divided by total number of tests).
10. Proceed to the next critter and repeat.
Cull: examine all the living critters on the grid and gather the top ten percent based on efficiency, and terminate all the others.
Repopulate: for the most efficient critters, allow each one to reproduce asexually (possibly with mutation). If after processing them all, the grid is not filled, go back and run through them again. Keep doing this until the grid is filled.
Loop: Jump to step 3.

An immediate problem which cropped up was that Weelanders would mutate in ways that made them faster at factoring this specific set of test numbers. So if by random chance, none of the test numbers was a multiple of say, 7, the Weelanders might make themselves faster by throwing away the "divide by 7" test somehow. Which is great until you hand them a multiple of 7 at which point they die from inaccuracy.

I solved this problem somewhat by changing step 6 to jump to step 2 instead of 3. Thus on each "tick" of the world, a new set of test numbers would be created and then all the Weelanders would be tested against that set. Makes sense really, and automatically weeds out Weelanders that made the grade last tick by "cheating". I improved this further by making the first 10 test numbers a fixed set that didn't change and which exercised a lot of the conditions I wanted to make sure were covered. The remaining 40 were random.

But another problem appeared, one that completely flummoxed me because the system appeared to drift toward less efficient creatures over time, instead of more efficient ones. How could that be possible? Natural selection should continue to hold true, and the creatures should be *more* efficient over time, not less efficient.

But nothing prepared me for the worst problem at all... hardware. My computer, wonderful though it is, is experiencing a hardware problem. The cooling system is no longer operating, or not operating well. As a result, any operation that pins the CPU usage at 100% for a period longer than 10 minutes or so will cause an audible heat warning alarm to begin issuing. It's an alternating two-tone klaxon which doesn't come from the computer speakers, it comes from somewhere inside the computer case. And it is, needless to say, very alarming. I like my CPU. I don't want to cook it.

Needless to say factoring numbers is CPU/FPU intensive. And I've got 230 Weelanders factoring 50 numbers each, over and over, ad infinitum. After 10 minutes or so of runtime, the CPU gets too hot, and I have to shut the simulation down or risk damaging my computer. This problem cropped up immediately after building the new GeneticFactoring code and caused me to terminate the experiment permanently months ago.

Then last night I broke the code out again and tried to figure out a way to get it to execute without pan-searing my CPU. Eventually I came up with a rather simple solution. The simulation executes for 20 ticks, pinning the CPU at 100% for about 60-90 seconds. Then the simulation pauses, basically issuing a call to Thread.sleep for sixty seconds of downtime. This drops CPU consumption to 0% or close to 0% for a minute, giving the CPU a chance to cool. Resulting in a usage pattern like this:

I'm happy to say that having made this change I've been running the simulation for 3 hours with no overheat alarms. At this point I feel safe walking away from the computer and letting it process the simulation.

Regarding the other problem, the problem of the population drifting toward inefficiency, I also managed to solve that. I observed the changes to the Weelander genome over time. After awhile I began to observe what I thought was the culprit. Creatures weren't being selected based on their efficiency at factoring *all* numbers, just the 50 test numbers they had to factor.

So let's say a mutant does something weird like tests factors out of order... instead of 2, 3, 5, 7, 11 it goes 5, 3, 2, 7, 11. It's still accurate, but it tests a different factor first. If more of the numbers in the test set are multiples of 5 than 3, and more are multiples of 3 than 2, then this critter will in the end rate more efficient than the basic Facto-f genome, which is suitable for factoring any value. As a result, genomes which aren't really more efficient at factoring any value, get selected ahead of the f genome, and over time, this exterminates the f genome. After which the population becomes less efficient over time as more and more test sets are produced.

That was my hypothesis anyway. To solve the problem, I decided to artificially inject the original genome into the list of critters during the repopulate stage, thus if the population ever drifted away from efficiency, there would always be some number of the original genome to tug it back. Basically I grab my ten percent most efficient, and then tack 10 copies of the original genome to that list before I repopulate. This makes sense since natural selection is only concerned about the environment you are in (i.e. the 50 numbers you happened to test) as opposed to my overall goal (more efficient factoring of ANY number.) Now the top ten percent must always compete with the baseline. If they drift away, the baseline will win and it will get selected for repopulation instead. The only way to stay alive is to consistently beat the baseline.

Makes sense doesn't it? Yes, yes, I know, brilliant.

It was right about that time I noticed that the code that was selecting genomes for propagation was selecting the least efficient ones instead of the most efficient ones. Y'know, it takes a special kind of mind to come up with a brilliant explanation like the one above and still be completely wrong. Needless to say selecting the least efficient members of the population will uh, tend to make the population drift toward inefficiency. It's amazing what happens to the code when you say "greater than" but you meant to say "less than". > versus <, baby... classic coding error.

Nonetheless, although the effect I hypothesized was ultimately not to blame for the drift, I still think the effect could occur, so I decided to keep the "baseline infusion" code in place.

So the factoring Weelanders are back in action, and I'm interested to see what they will do to become more efficient at factoring. There are a lot of simple things I can think of that would lead to slightly more efficient code, but the Weelanders are always surprising me.

What I can tell you is that given a test sample, the Facto-f genome generally requires (on average) 1500 to 2200 execution steps per number. After running the simulation for 3 hours, the top genomes are showing efficiencies on the order of 500 to 600 execution steps per number. Ok I'm impressed. Especially when you consider that these Weelanders have to produce accurate results to survive. If they fail any single factoring test, they're gone.

I'm going to let the sim run for awhile longer and then take a peek at these efficient Weelanders. I wonder what they are doing?

Beware of the Bob

Abacquer — Thu, 19 Oct 2006 21:19:17 -0400

So anyway, I'm going to talk more about Weelanders. You may be tired of hearing about my little artificial life forms by now but then, that's your problem. Today I want to talk more about the (mistaken) belief that much of life is either too perfectly suited to its environment or too complicated to have evolved through random mutation and natural selection. To illustrate this, I will use Weelander genome i2132 (who I have nicknamed "Bob"), who is one interesting fellow...

Weelander Explosion

Abacquer — Wed, 18 Oct 2006 00:15:42 -0400

Since giving Weelanders the ability to have sex a number of interesting things have happened...

Sex Wins

Abacquer — Mon, 16 Oct 2006 01:26:01 -0400

So over the weekend the Weelanders finally acquired the means to share genetic matter, and sharing genetic matter seems to be a more reliable way to create a healthy population...

Weelanders in Action

Abacquer — Sat, 14 Oct 2006 13:33:02 -0400

WARNING: this article is long!

Okay, more on Weelanders, my latest point of interest, including some output for you to peruse. Here's a run I started with a slightly modified Basic-c genome (called Basic-c2). I made a change based on something the Weelanders taught me about my food search algorithm...

I've Got a Secret

Abacquer — Fri, 13 Oct 2006 16:11:21 -0400

Awhile back I mentioned I was using Visual Studio, .NET, and C# to conduct an experiment I have always wanted to run, as a vehicle for learning about .NET. I also refused at the time to reveal the nature of it because I was afraid I wouldn't follow through. Well I followed through, and it is beginning to bear fruit, and is, in my opinion freaking cool...

It's 2006...

Abacquer — Mon, 24 Jul 2006 11:39:51 -0400

And I'm downloading indie music for my MP3 player from a picnic table in a wide courtyard surrounded by tall pines, in the shade of an umbrella with a beautiful blue sky dotted by slow moving cumulous clouds above. The scent of fresh pine is on the breeze and I can hear little kids playing in the distance. Damn. Technology is something... the future is now.

(It's also Monday and I'm not even thinking about work. That's the effing bomb, yo!)

Something Beautiful

Abacquer — Fri, 30 Jun 2006 20:12:41 -0400

Go and watch this movie, and keep in mind it's 100% real, and was shot entirely in 2006. There's an earlier film on the site that was put together in 2005 that follows the same format but both films amazed me...

Stone Calendar?

Abacquer — Wed, 28 Jun 2006 11:46:48 -0400

I love stone calendars like Stonehenge--rings of stones marking days of the year or important annual astronomical alignments. Such structures are common in many parts of the world, but until now, none have been found that originate from pre-Columbrian Amazon cultures, simply because such cultures tended to build using clay and wood. Now, on a grassy hilltop near the village of Calcoene in northern Brazil, scientists have discovered what may be the remains of a stone calendar...

Methuselah's Baby

Abacquer — Tue, 13 Dec 2005 09:52:43 -0500

Members of the Champion Tree Project International recently bequeathed, as a Christmas gift, to the Washington National Cathedral, a seedling. Giving the gift of a seedling is cute, but what makes this seedling so special is that it germinated from a pinecone dropped by "Methuselah", a particular bristlecone pine tree growing at an altitude of 11,200 feet in the White Mountains near the California-Nevada border. This tree currently holds the distinction of being the oldest known living tree in the world at 4,770 years. Cool! (Yahoo News Story)

Hayabusa's Packing Asteroid

Abacquer — Sat, 26 Nov 2005 09:05:26 -0500

From Japanese Probe Collects Asteroid Samples (AP via Yahoo News):

..."The initial movements and operations look very good," Yashiro said. "The process of sampling also seems to have gone very well...

...But JAXA won't know if Hayabusa actually collected surface samples until it returns to Earth. It is expected to touch down in the Australian Outback in June 2007...

...Hayabusa touched down shortly before 8 a.m. Japan time, firing a metal projectile into the surface and collecting the dust that was kicked up. The whole procedure was over in a matter of seconds...

Initial Hayabusa Touchdown Confirmed

Abacquer — Thu, 24 Nov 2005 22:32:44 -0500

During the three hours of radio silence during Hayabusa's touchdown attempt the other day, the craft did indeed land on the Itokawa asteroid according to JAXA officials. It did not however collect a sample as it was supposed to. Another landing attempt is scheduled for Saturday. Hayabusa's Earth-return window closes in early December, so I imagine the science team is feeling a little pressed for time at this point!

Next Stop: Venus!

Abacquer — Wed, 23 Nov 2005 16:00:00 -0500

The European Space Agency's (ESA) Mars orbiter (the Mars Express) has been a great success. Its mission is expected to wrap up on November 30 of this year. On November 9, ESA launched a sister mission to Venus... the Venus Express! The ship is expected to arrive at Venus in about 4 months where it will orbit the hot planet and study its soupy, hostile atmosphere. Check out the Venus Express Mission Homepage for more information.

Another interesting upcoming ESA mission (due to launch in June of 2006) is COROT. This satellite will orbit the Earth and will include equipment which should be able to definitively detect terrestrial planets (i.e. consisting largely of rock, as opposed to gas giants) around other stars. Up until now the only planets we've been able to detect around other stars have been gas giants like Jupiter, the largest and most massive sorts of planets.

Pluto and Charon Have Children

Abacquer — Fri, 18 Nov 2005 10:31:22 -0500

Pluto, the enigmatic little Kuiper Belt object (or planet, if you prefer) which wanders in the far reaches of our solar system, has long captured my imagination. As an 11 year old boy, I was thrilled with the news that a moon of Pluto, "Charon" had been discovered. I liked the idea that Pluto had an inordinately large (by comparison) companion, that it didn't have to wander the cold frontier alone. Then earlier this month, I got to relive that thrill...

Happy Birthday James!

Abacquer — Sun, 30 Oct 2005 08:18:48 -0500

Please join me today in wishing my good friend James a happy birthday! James is 38 today, I think.

I met him in college approximately 20 years ago and I've been all the luckier for it ever since.

So head on over to Aces Full of Links and wish him a happy day!

List of Fallacies

Abacquer — Tue, 25 Oct 2005 16:28:58 -0400

I found this today while tooling about the 'net. It's a list of common fallacious arguing tactics with explanations. Neat!

Kid Catches 5 Foot Bass, Wins Competition

Abacquer — Wed, 19 Oct 2005 11:14:38 -0400

This is a cute story, a 12-year-old girl with the prophetic name Molly Fischer showed just what a fisher she is this year by winning the Martha's Vineyard striped bass fishing tournament with a 5 foot 1 inch, 49.2 pound bass she caught...