[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 95 KB, 1280x720, Spider (2002).jpg [View same] [iqdb] [saucenao] [google]
5701524 No.5701524 [Reply] [Original]

What are some problems that Genetic Algorithms are best suited to solve? (Or at least most often used for)

>> No.5701527

What is the source on your pic?

>> No.5701531

Optimization problems. Minimizing/maximizing functions, training fuzzy logics, finding parameters for control schemes....

>> No.5701534

>>5701527
It's in the file name, it's a film called "Spider" from the year 2002. That is actor Ralph Fiennes.

>> No.5701541

>>5701531
What's the benefit compared to traditional curve fitting?

>> No.5701582

What are they like at solving equations?

Are they pointless because the fitness function basically means they need to already know the answer to get the answer?

>> No.5701584

>>5701541
faster for some things.

>>5701582
you dont know the answer if you know the fitness function, lets say you use it to find the 0s of x+x^2. then you dont need to know the answer, you minimize (x+x^2)^2

>> No.5701598

anything that's a white noise space.

>> No.5701622

I used one that could generate StarCraft build orders.

You told it what you wanted, and it would find an optimal build via genetic modification.

>> No.5701627

>>5701622
It actually fucked the meta game for awhile, by generating the 7 roach rush, a build that got roaches out insanely fast.

>> No.5701629

>>5701627
i remember that, you could only counter it if you knew someone was doing it and you were protos.

>> No.5701634

>>5701524
not really what you asked but stuff like this is cool
http://rogeralsing.com/2008/12/07/genetic-programming-evolution-of-mona-lisa/

>> No.5701656

The Travelling Salesman is a classical AI problem, does it fair well on that?

>>5701598
Could you elaborate?

>>5701634
That is cool!

>> No.5701751

Solving NP-complete problems that have good Heuristics in general?

>> No.5702203

>>5701584
>faster for some things.
Which things are they good at?

>> No.5702654

Are GAs as biologically inspired as say spiking neural nets or are they really toned down simplifications?

>> No.5703305

>>5702654

well, they just use the principle of natural selection. if it is "fitter" (for purpose) then it survives. it's very hard to simplify evolution any more than it already is. specific mechanisms are almost irrelevent as long as the principle of selection remains.

so i wouldn't really call them simplifications, although they can be very simple.

>> No.5703910

>>5703305
>specific mechanisms are almost irrelevent as long as the principle of selection remains.
WUT

>> No.5703948

>>5701524
GAs are great for optimization problems where the curve has a very complicated or unknown shape or has a lot of local minima/maxima. It's a randomized algorithm, and doesn't need to know anything about what it optimizes. The downsize, of course, is that the running time can be pretty slow.

If you're trying to optimize something that has a simple shape, like in a convex optimization, then there are much better techniques that can utilize the convexity to give optimal or near optimal answer in a shorter time period.