Slashdot Mirror
The Game Theory of Life
An anonymous reader writes In what appears to be the first study of its kind, computer scientists report that an algorithm discovered more than 50 years ago in game theory and now widely used in machine learning is mathematically identical to the equations used to describe the distribution of genes within a population of organisms. Researchers may be able to use the algorithm, which is surprisingly simple and powerful, to better understand how natural selection works and how populations maintain their genetic diversity.
1. If the machine learning algorithm has been found to be mathematically identical to the genetic spread algorithm, how would biologists be able to use it to better understand natural selection and genetic diversity? What can they learn from the first algorithm that they couldn't learn from the one they already had? If the two algorithms are mathematically identical, aren't they both just different names for the same mathematical structure? Learning a cat is called neko in Japanese doesn't tell you anything about cats you didn't already know -- it just tells you something about the Japanese language.
2. Are algorithms discovered, or created? If anything is discovered, the underlying mathematical structure more than one algorithm can point to seems to be a better candidate than the algorithms themselves. Fossils are discovered; algorithms are made up.
...has the "simulated universe" hypotesis just got a slight boost from this finding?
Knowledge is power; knowledge shared is power lost.
Our Learning is a combination of what we associate with the result vs. what we experience the first time. However, we cannot be entirely sure of the result unless we experience it multiple times;
This forces us to create biases, learnings and mind sets that somehow seemingly protect us and help us survive - while driving us irrational most of the time.
Neural network algorithms most certainly operate in the same way.
Sounds the same as conclusions by researchers cited in Matt Ridley's book "The Origin of Virtue".
Only having read the abstract, and the linked article, I don't really see how this is different from the "2 Armed Bandit" theory which John Holland Laid out 40 years ago in "Evolution in Natural and Artificial Systems". Holland laid out how the combination of sexual reproduction with mutation within a population otpimises search across the space by combining exploitation of good areas of the search space with exploration to find better areas.
Can someone more up to date enlighten me?
kind regards
tree frog
"equations used to describe the distribution of genes", meaning the algorithm (mechanism) was probably unknown while the equations just described the end result. Also, the behavior of an algorithm that sees lots of use has probably been studied a lot so there's a treasure trove of material for researchers to dig through to find stuff relevant for genetics.
Things in mathematics can be independently verified. You can give a theorem to a computer and have it search for a proof. The proof either exists or it does not. You can even prove it is unproveable. You can't have a computer generate random words and have it know when its quoted Shakespeare (without a reference copy to check against).
So maths is more like the world, a country (proof) exists whether you have discovered it yet or not. The same cannot be said of Shakespeare.
It has been known for years, probably decades, that gene frequencies follow this mathematical rule, and that it has been mathematically proven optimal for solving Multi-armed bandit type problems. Each generation genes are tested by natural selection, and increase or decrease in frequency according to multiplicative increase or decrease. This is a mathematically optimal strategy for exploring and optimizing payoff in a complex unknown environment. Mutation creates random stuff to try, and this mathematically selection algorithm optimally crafts it into useful new information.
-
- - You can't take something off the Internet! That's like trying to take pee out of a swimming pool.
Why does Slashdot seem to buy in so often to spinning the recurrence of mathematical tools across various fields as some kind of scientific breakthrough? Correlation is not causation, not all structural similarities imply some kind of necessary physical theoretical account. We as empirical agents use logical tools for the formation, quantification and application of theories - so of course some functions will occur in several different settings, because we're bringing the same resources to the table each time.
Myu:
The Earth was created 7000 years ago by some unknown but benign power, evolution doesn't exist - how can some equation possibly describe it using mathematics? Mathematics itself is obviously a made up tool anyway, so you can make it say whatever you want. They'll be using Empirical Science next to debunk what is clearly the One Truth. :)
"Is the Chief Priest an Offlian? Do dragons explode in the wood?"
That's the first thing I thought also. This isn't the first time someone has thought about this problem.
... I imagined that someone discovered a mechanism in genes that favored survival if there were exactly two or three neighboring genes, and non-survival if there were fewer or greater numbers. Oh, and something about a new gene being 'created' when there are exactly 3 neighboring ones.
- Mike
Drifting off topic, but did the infamous Beta in fact get Unicode support?
I mean, look at this tortuous new Beta, did they even bother to put in the Unicode support that people have been screaming for for ten+ years?
Damn we need a mole at Dice. What do they even do at management meetings?
"Let's make a whole new design with 55 changes."
"What about Unicode Support?"
"That's a big word. That's too hard for me. Let's put more videos up instead."
My first Journal Entry ever, in 8 years! http://slashdot.org/journal/365947/aphelion-scifi-fantasy-horror-poetry-webzine
John Maynard Smith introduced the game theory to evolutionary biology in the early 70's. It was a breakthrough at that time, however today it is scarcely news. Evolutionary biology, and in especially population genetics has been a highly mathematized discipline ever since before WWII, when it was developed by Fisher, Wright and Haldane. Later you had Hamilton and Maynard Smith. It is nice that computer scientists noticed that something exciting is going on here, but don't fall for press releases and insubstantiated claims.
"We demonstrate that in the regime of weak selection, the standard equations of population genetics describing natural selection in the presence of sex become identical to those of a repeated game between genes played according to multiplicative weight updates (MWUA)..."
There's a preprint of what seems to be a more complete paper on the work hosted on arxiv. There's a bit more math in it, but it's still somewhat understandable: http://arxiv.org/pdf/1208.3160.pdf.
That is all.
There isn't an algorithm as the world was created 6000 years ago out of nothing. Of course maybe the mystical sky daddy also created the algorithm too in order to test your faith in ghosts of the holy kind. Maybe we should call that ghost Casper just to make him more approachable.
beta does not have unicode support