View allAll Photos Tagged algorithmic
I decided to try out neural nets and genetic algorithms at once to see whether I could get them to work.
A genetic algorithm has a system somewhat inspired by our own DNA. You construct a chromosome of values, then you have some function to interpret the values and get an answer, and a fitness function to measure how correct the answer is. You have the chromosomes reproduce by combining according to some rules for the next iteration (giving preference to ones that did better in the last iteration, etc) and with some chance of mutation, and so forth. Over bunches of iterations, successful chromosomes take over.
Neural nets are inspired by the way neurons in our brains work. Each neuron has a bunch of inputs and weights tied to each input, and a bias value... And the neuron fires or not based on the sum of the inputs times their weights minus the bias. You can change neurons from on-off to a smoother response via a sigmoid function.
I made a simple game where red balls are trying to eat stationary blue balls. The neural net controls the acceleration of the red balls based on current velocity and the location of the nearest ball. The chromosomes contains the weights and bias values for the neural net. Early on, the chromosomes are random gibberish so the balls would often just fly off in a certain direction. After a while, they started adjusting their path to hit the blue balls. After 1500 iterations they all exhibited fairly complex and much more successful behaviors despite the very simple neural net.
I assumed it'd figure out to aim for the closet blue ball but that's not what happened. The strategy that took over consisted of getting up to top speed, totally ignoring the direction of the nearest ball until it happens to get really close. Then it'd swerve sharply and pick it up. I had wraparound on, so balls leaving the left side show up on the right, etc. so that helps the strategy. I'm going to turn off wraparound so they can drive themselves into a corner uselessly and see if it evolves the strategy I expected.
Repeated application of the Baker's algorithm (stretch, fold, turn, repeat) leads to some colorful artifacts.
Algorithmic art created with Processing using blue noise dot patterns, wave functions, and various other computational techniques.
Shift Scan, 89 x 256 pixel gaps between 512, 1024, 2048, 4096 and 8192 pixel shapes mapped to a HIlbert Curve.
The Algorithm (métal progressif / electro, France), festival Prog Frog, 31 mars 2017, KIFF, Aarau (Suisse).
Photo: Stéphane Gallay, sous licence Creative Commons (CC-BY)
An algorithm, in computer science and mathematics, is a self-reliant gradual set of operations that is to be executed. There are algorithms that can efficiently perform data processing, automated reasoning, and calculations. It is a very effective method for calculating a function, which can be...
www.technowize.com/algorithm-surpasses-human-intuition-a-...
The Algorithm (synthwave, France) en concert aux Docks de Lausanne, le 9 octobre 2019.
Photo: Stéphane Gallay, sous licence Creative Commons (CC-BY)
Shuffle technique that works for arbitrary bit lengths, and slowly increases in scale from the left to right.
A zoomable version can be found here.
This is a picture of a "dense" Julia set. More information can be found here.