Neural network in GML!
gms 2.3 abilities!
simple single-layer perceptron
easy way to train neural network with limit in iterations and error level
save and load networks
nn_create() - create neural netork, return id
nn_destroy(id) - destroy neural network
nn_add_neuron(id) - create new neuron in current network
nn_get_output(id, input_list) - launch neural network and return list with output
nn_train_train(id, train_id) - train neural network
nn_get_neuron_number(id) - return number of neurons in network
train_create() - create information for training neural networks, return train_id
train_add_input(train_id, val, val, ...) or train_add_input_list(train_id, input_list) - add input signals for training
train_add_output(train_id, val, val, ...) or train_add_output_list(train_id, output_list) - add desired signals for comparison
train_destroy(id) - destroy current train
You can now create neural network struct using
network = new neural_network(n); !
And use this methods:
.. other functions from list above ^ ...
buffer = network.save_to_buffer()
Some examples inside.
This is simple single-layer perceptron, I hope this example will help many people to understand that neural network is easy.
( Don't forget to rate :D )
Implemented 2.3 features:
Neural network wrapper using constructors: net = new neural_network(...);
All functions in one file