Backpropagation neural network

neural-network is a library that allows you to create and train a neural network of a given configuration.

Create network

create_network(layers_size)

• layers_size - list of the number of neurons on each layer (starting from the input layer)

Train network

train(weights, test_datasets, iterations, learning_rate)

• weights - network weights
• test_datasets - list of typles of input datasets and expected datasets
• iterations - number of interations
• learning_rate - (0, 1) neural network learning speed / accuracy

Predict

predict(weights, input_datasets, reliable_limit)

• weights - network weights
• input_datasets - list of input datasets
• reliable_limit - (0, 0.5) limit of unknown results

Example:

train_datasets = [
        ([0, 0, 1], [1]),
        ([0, 1, 1], [0]),
        ([1, 0, 1], [1]),
        ([1, 1, 1], [1]),
        ([0, 0, 0], [0])
        ]

weights = create_weights([3, 2, 1])
train(weights, train_datasets, 6000, 0.05)
assert predict(weights, [0, 1, 0], 0.35) == [0]

Contributors

• See github for full contributors list