MetAISmith / myai

Create some ai use the BugBrain to play openai gym

myai

This program uses the BugBrain to build AIs to play in the OpenAi gym.

Introduction

BugBrain

• This neuron network comes from the game 'BugBrain'. Import BugBrain to use it.

    import BugBrain as BB

• Use class Brain to build a brain

    brain = BB.Brain()

• Then add neurons to the barin. The neuron has three different activation functions. They are 'Step', 'Linear', and 'Sigmoid'.

    brain.neurons.append(BB.Neuron('Step'))

• Create a inputnode and a synapse to the neuron.

    inputnode = BB.InputNode()
    brain.neurons[0].synapses.append(BB.Synapse(inputnode))

• In this way there will be a link from the inputnode to the neuron. Then you can change the weight of the synapse or the bias of the neuron. Set the inputnode and run the barin.

    inputnode.value = 1
    brain.work()
    output = brain.neurons[0].value

worm

• Now the worm is used to play the 'CartPole-v1'. There is only one neuron in its brain. I use evolutionary algorithms to train the worms.
• I am trying to enable them to evolute their brain's shape automatically, and then they can play more difficult games.
• The BugBrain has a huge potential. Next time I will try to enable the decay function. Then the worms can remember things.

How to Use

• install gym
• run 'worm.py'
• you can see after many generations the worm can play well.

Environments need

gym