Two Neurons Worm

Two Neurons Worm

Elapsed:

Generation:

Eaten:

Iterations Left:

Generation Settings

show

Apples:

Worms:

Maximum Iterations:

Speed:

Mutation level:

Random worms per generation (%):

Top Worms

show
# Worm Belly
1
2
3
4
5

Top Ancestors

show
# Family Gen Breed
1
2
3
4
5

Eaten Graph (%)  (a/w)

show toggle
-WAITING FOR DATA-

Worms Brains

show

Algorithm:

description

In the Random Walk algorithm, the worm has a chance of changing it's dicrection at each iteration.

The new random worms are sampled from a gaussian distribution and each have it's own chance of turning according to the mean and standard deviation in the sliders below.

The breed of the most successful worms also changes it's chance of turning according to the Mutation level defined in the Generation Settings.

Chance of turning - mean (%):

Chance of turning - std (%):

description

The neural Two Neurons algorithm calculates if the worm should turn based on the smell of the apples. By giving the previous smell as input, the worm can detect when it is going closer to an apple and when it is going away.

It's possible to chose the activation function for each neuron. The output of the first neuron dictates the size of the step. The output of the second neuron makes the worm turn if it's activated (output greater than 0.5).

The new random worms are sampled from a gaussian distribution and each have it's own set of neurons that can be configured below.

The breed of the most successful worms also changes it's neurons according to the Mutation level defined in the Generation Settings.

Neuron 1 activation function

Neuron 1 - mean:

Neuron 1 - std:

Neuron 2 activation function

Neuron 2 - mean:

Neuron 2 - std:

Worms Bodies

show
description

At each iteration, a worm can move forward or change it's direction.

The forward movement has an upper limit that is defined by it's maximum step size. New worms are sampled from a gaussian distribution for it's maximum step size.

The worm's turning is limited by it's turning range in degrees. There's also a preferencial bias for turning. Both parameters are gaussian.

Maximum step size - mean:

Maximum step size - std:

Maximum turning range - mean (°):

Maximum turning range - std (°):

Maximum turning bias - mean (°):

Maximum turning bias - std (°):

Inspect Worm

show

click in a worm or in the Top Worms table

Name

Belly

Algorithm

Turning chance (%):

Neuron 1 - Step forward

Bias:

Input 1 (scent):

Input 2 (previous scent):

Input 3 (turned last iter):

Neuron 2 - Turn

Bias:

Input 1 (scent):

Input 2 (previous scent):

Input 3 (turned last iter):

Body

Maximum step:

Max left turn:

Max right turn:

Check it on GitHub!

Copyright (c) 2017 Luiz Eduardo Amaral - <luizamaral306@gmail.com>

This software is under a GNU GPL3 license. Have fun! 😉