HackerLastPip’s favorites
Recently commented on
genetic algorithms: pecha kucha style
Project: Emergence and Navigating Space
Project: Emergence and Navigating Space
I did a presentation on Genetic Algorithms for programming media I several weeks ago in the pecha kucha style and it turned out pretty well. In keeping with the idea, I'm going to write the descriptions to these images in 20 seconds per slide to give you a brief overview on the subject. EDIT: I had to have myself 40 seconds per slide... I type much slower than I talk.
Ok this slide pretty much describe the whole process; you have a bunch of bots/ virtual creatures, they mate, they are judged on their fitness, and they are eventually killed, it's just like life / Darwinism which is was based
This diagrams exactly how it happens, 1 you have a GOAL, that is what drives the whole process, 2 you select ones that are closer to the goal, 3 breed them and select their offspring leading to a GRADUAL change
You can use these genetic algorithms to breed/develop all sourts of things from virtual creatures (Calrl Sims), to
paths, to placements of chess boards to the fractal patters in the background
This figures shows why you need to select digital organisms/ things that are not perfect or even remotely perfect ( freaks). they may, even though they are different, have that hidden key things that leads you to the goal
This is from the Electric Sheep project, and shows how this concept of genetic breeding works great for aesthetic things when you harness the internet and have millions of people voting, to replace your evaluative algorithms
A fun picture illustrating how you must must must rate fitness. you have to choose what's good and what's bad, or else it wont work. algorithms that separate the good from the bad are much of the hard part
To show the power of this technique a computer scientist set a herd of virtual, replicating bots on his computer and before long there were creatures who were half the size theorized possible moving around, who had evolved innovations unknown.
This illustrates the phenomenon of falling in a ditch so to speak. your goal is on the hill , but here's a dip, so any new creatures are less fit by moving towards the goal, so you end up with a bunch on non fits
This image is from the electric sheep project. All of these images were created by people voting on algorithms they though were appealing. The most fit ones (judged by votes) survived and mated.
Ok this slide pretty much describe the whole process; you have a bunch of bots/ virtual creatures, they mate, they are judged on their fitness, and they are eventually killed, it's just like life / Darwinism which is was basedThis diagrams exactly how it happens, 1 you have a GOAL, that is what drives the whole process, 2 you select ones that are closer to the goal, 3 breed them and select their offspring leading to a GRADUAL changeYou can use these genetic algorithms to breed/develop all sourts of things from virtual creatures (Calrl Sims), topaths, to placements of chess boards to the fractal patters in the background
This figures shows why you need to select digital organisms/ things that are not perfect or even remotely perfect ( freaks). they may, even though they are different, have that hidden key things that leads you to the goalThis is from the Electric Sheep project, and shows how this concept of genetic breeding works great for aesthetic things when you harness the internet and have millions of people voting, to replace your evaluative algorithmsA fun picture illustrating how you must must must rate fitness. you have to choose what's good and what's bad, or else it wont work. algorithms that separate the good from the bad are much of the hard partTo show the power of this technique a computer scientist set a herd of virtual, replicating bots on his computer and before long there were creatures who were half the size theorized possible moving around, who had evolved innovations unknown. This illustrates the phenomenon of falling in a ditch so to speak. your goal is on the hill , but here's a dip, so any new creatures are less fit by moving towards the goal, so you end up with a bunch on non fitsThis image is from the electric sheep project. All of these images were created by people voting on algorithms they though were appealing. The most fit ones (judged by votes) survived and mated. 
Add comment
| Promote |
|
Add to favorites
|
| Create synapse |
|
