mutationNN creates a Neural Network and allows you to mutate it.

How to use:

from mutationNN import NN

1. Create a NN

neural_network = NN(

INPUT_SIZE = 3,

OUTPUT_SIZE = 1,

MUTATION_CHANCE = 0.25, #0.25 is the default setting.

FULLY_CONNECTED = True #False is the default setting.

)

2.1. The Output of the NN will Change the next Input:

p_I = [1, 2, 3]

output = neural_network.process_Input(p_I) #Gives you a List with the Size OUTPUT_SIZE

2.2. The Output won't Change the next Input:

p_S_I = [

[1, 2, 3],

[2, 3, 4],

[3, 4, 5]

]

output = neural_network.process_Static_Input(p_S_I))

3. Mutate the NN:

neural_network.mutate(must_mutate=False)

# must_mutate has False as default but can be set to true in order to ensure at least 1 mutation will happen.

4. Take another NN as template:

neural_network2 = NN(

INPUT_SIZE = 3,

OUTPUT_SIZE = 1,

MUTATION_CHANCE = 0.25, #0.25 is the default setting.

FULLY_CONNECTED = True #False is the default setting.

)

neural_network.copy_NN(neural_network2)

# Will copy the Neurons and Weights of agent_2 and apply mutate(must_mutate=True) to the network afterwards.

5. Save your NN:

neural_network.save_NN(path="Here your Path", name="NN")

# NN is the Default Name. It is a .json file.

6. Load your NN:

neural_network3 = NN(

INPUT_SIZE = 0,

OUTPUT_SIZE = 0

)

neural_network3.load_NN(path="Here your Path", name="NN")

# INPUT_SIZE/OUTPUT_SIZE and MUTATION_CHANCE don't matter since all Information of the Initial

NN will be overwritten by the loaded NN.

# NN is the Default Name.