Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics


Create creatures and let them evolve to see how they master various tasks. · By Keiwan


A topic by Ericenjoy created May 09, 2018 Views: 975 Replies: 3
Viewing posts 1 to 4

I have no idea how the brain works, does anyone know? can you try to explain it to me?

Yes, it would be really interesting to know.


(2 edits) (+2)

The neural network uses weighted connections and neurons to simulate biological brain processes. each input is a run through a  funtion to generate a sigmoid curve, basically the input can be between 0 and 1 so the creature does not get an input of something crazy and have a seizure. Each input is multiplied by the weight of every connection (-1 to 1) and that number is added to a total of the neuron at the end of the connection in the next layer forward, once again the sigmoid is applied to normalise the result to between 0 and 1 and this continues until the final value where the output is calculated, the NN mutates changes in the weights of each connection to provide different outputs. For example, if the foot touches the floor the input for that joint will be 1, and the weight to the output muscle expansion will be maximum, so if the foot touches, then the muscle will have a strong output reaction rather than a small one, if the weight is negative then it inhibits the output so if the foot is in the air, then that will stop the muscle from activation with a strong negative weight. Each neuron also adds a bias to its final output so the result can be within a different range, if the normal range is 0 to 1 adding a 5 bias will lead to a number between 5 and 6 which helps to activate some outputs more strongly. That is basically how every Neural network works, so im pretty sure it is accurate to this simulation.


Input, connection, Output

1 ------ -1 -------1 :  1 * -1 = -1
0 ----0.5 ------ 0 : 0 * 0.5 = 0

if input a is a foot touching the floor, output a is the leg muscle expanding.