class MNISTTrainer(object):
def __init__(self, gui_worker=None):
print('----> Loading cases...')
training_set = load_all_flat_cases('training')
testing_set = load_all_flat_cases('testing')
provided_datasets = [
training_set,
testing_set
]
self.net = Ann(
structure=[784, 200, 40, 10],
datasets=provided_datasets,
activation_function=[T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid],
learning_rate=0.2,
regression_layer=SumOfSquaredErrors,
gui_worker=gui_worker
)
self.gui_worker = gui_worker
def train(self, epochs=80):
self.net.train(epochs)
def test(self):
minor_demo(self.net)
def __init__(self, structure=None, provided_datasets=None,
activation_function=None, regression_layer=None,
learning_rate=None, gui_worker=None):
if not provided_datasets:
min_tile = None
heuristic = None
num_cases = None
if gui_worker:
min_tile = gui_worker.gui.min_save_tiles
heuristic = gui_worker.gui.heuristic
num_cases = gui_worker.gui.num_cases
else:
print('----> Loading cases...')
training_set, testing_set = load_2048_example(min_tile, heuristic, num_cases)
provided_datasets = [
training_set,
testing_set
]
self.net = Ann(
gui_worker=gui_worker,
structure=structure if structure else [17, 300, 60, 4],
datasets=provided_datasets,
activation_function=activation_function if activation_function else [T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid],
learning_rate=learning_rate,
regression_layer=regression_layer,
normalize=process_states
)
self.gui_worker = gui_worker
class Ann2048Trainer(object):
def __init__(self, structure=None, provided_datasets=None,
activation_function=None, regression_layer=None,
learning_rate=None, gui_worker=None):
if not provided_datasets:
min_tile = None
heuristic = None
num_cases = None
if gui_worker:
min_tile = gui_worker.gui.min_save_tiles
heuristic = gui_worker.gui.heuristic
num_cases = gui_worker.gui.num_cases
else:
print('----> Loading cases...')
training_set, testing_set = load_2048_example(min_tile, heuristic, num_cases)
provided_datasets = [
training_set,
testing_set
]
self.net = Ann(
gui_worker=gui_worker,
structure=structure if structure else [17, 300, 60, 4],
datasets=provided_datasets,
activation_function=activation_function if activation_function else [T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid],
learning_rate=learning_rate,
regression_layer=regression_layer,
normalize=process_states
)
self.gui_worker = gui_worker
def save(self, path=None):
if not path:
path = res.play2048s.anns.__path__[0] + "/trained-ann-2048"
file = open(path, 'wb')
pickle.dump(self.net, file, protocol=pickle.HIGHEST_PROTOCOL)
file.close()
def train(self, epochs=100):
self.net.train(epochs)
def __init__(self, gui_worker=None):
print('----> Loading cases...')
training_set = load_all_flat_cases('training')
testing_set = load_all_flat_cases('testing')
provided_datasets = [
training_set,
testing_set
]
self.net = Ann(
structure=[784, 200, 40, 10],
datasets=provided_datasets,
activation_function=[T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid],
learning_rate=0.2,
regression_layer=SumOfSquaredErrors,
gui_worker=gui_worker
)
self.gui_worker = gui_worker
if __name__ == '__main__':
print('----> Loading cases...')
training_set = load_all_flat_cases('training')
testing_set = load_all_flat_cases('testing')
provided_datasets = [
training_set,
testing_set
]
net1 = Ann(
structure=[784, 150, 10],
datasets=provided_datasets,
activation_function=[T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid],
learning_rate=0.1,
regression_layer=SumOfSquaredErrors
)
net1.train(100)
minor_demo(net1)
net2 = Ann(
structure=[784, 150, 10],
datasets=provided_datasets,
activation_function=[T.nnet.sigmoid, T.nnet.sigmoid, T.nnet.sigmoid],
learning_rate=0.2,
regression_layer=SumOfSquaredErrors
)