The Daily Click ::. Downloads ::. Miscellaneous ::. Behavioral Evolution Simulation

Behavioral Evolution Simulation
Author: Duckimation Submitted: 24th August, 2014 Favourites:0
Genre: Miscellaneous Downloads: 93

Edited By Duckimation on 24/08/2014

This simulation is supposed to simulate the evolutionary process with a genetic algorithm. Each creature thinks for itself and makes decisions based on its surroundings. Each brain is modeled by a neural network that takes inputs and passes values through nodes based on connection weights. Creatures follow the actions that result from the neural network function.
Creatures start off with random brains. Periodically, new random creatures are spawned to create more competition.
When a creature’s health decreases to zero from lack of eating food, it dies. When its health reaches a certain threshold, it spawns an offspring with a mutated copy of the brain, altering the behavior slightly. Evolution is based on survival of the fittest and descent with modification, and since creatures are competing for limited food to survive and reproduce, eventually an optimal behavior should evolve.

Creatures take data from food and other creatures near it as an input.
Turn speed and acceleration are given as outputs.
The faster a creature is moving, the faster health is consumed.
Creatures have a spike that they can extend to steal health from another creature.
They also can give off pheromones, or chemical signals, that other creatures can sense.
Creatures have a unique color that is inherited and can be sensed by other creatures.

The simulation can be paused at any time(p=pause, r=resume)
Creatures/food can be spawned/destroyed by the user(click+a=creature, click+d=food, click+x=kill)
A creature, with all of its neural data, can be saved and loaded into the simulation(click on a creature to select, “load” and “save” under “file” menu option)
The neural network can be configured to specify which senses/actions are to be used, as well as the size (complexity) of the hidden part of the network(under “options”> “configure”)
Lastly, the data of a selected creature can be viewed in a sidebar(“options”> “view properties”)

All this info is also available in the description section of the menu.

Review This Download (1.23 mkb )

Posted by Hagar 29th August, 2014

Very interesting, I will be front paging this shortly . How do you train you network weightings?
Posted by Duckimation 30th August, 2014

It's based on random mutation- creatures start off with random weights, and each offspring receives a neural network with the same weights, but altered slightly based on randomization. Evolution occurs because ones that are better at getting food have more offspring.
Posted by Billybobjoe198 9th October, 2016

The link is broken, dropbox doesn't like high traffic files.





Worth A Click