steffen/self-replicating-neural-networks
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Functional Interface

Browse Source
This commit is contained in:
Si11ium 2019-03-04 15:02:02 +01:00
parent b7f5747f50
commit 415c7e9d41
1 changed files with 23 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
code/network.py
View File

@ -17,6 +17,7 @@ def normalize_id(value, norm):
else: else:
return float(value) return float(value)
def are_weights_diverged(network_weights): def are_weights_diverged(network_weights):
for layer_id,layer in enumerate(network_weights): for layer_id,layer in enumerate(network_weights):
for cell_id,cell in enumerate(layer): for cell_id,cell in enumerate(layer):
@ -27,6 +28,7 @@ def are_weights_diverged(network_weights):
return True return True
return False return False
def are_weights_within(network_weights, lower_bound, upper_bound): def are_weights_within(network_weights, lower_bound, upper_bound):
for layer_id,layer in enumerate(network_weights): for layer_id,layer in enumerate(network_weights):
for cell_id,cell in enumerate(layer): for cell_id,cell in enumerate(layer):
@ -324,21 +326,30 @@ class RecurrentNeuralNetwork(NeuralNetwork):
class LearningNeuralNetwork(NeuralNetwork): class LearningNeuralNetwork(NeuralNetwork):
@staticmethod @staticmethod
def _apply_mean_reduction(self): def mean_reduction(weights, features):
return single_dim_weights = np.hstack([w.flatten() for w in weights])
shaped_weights = np.reshape(single_dim_weights, (1, features, -1))
x = np.mean(shaped_weights, axis=1)
return x
@staticmethod @staticmethod
def _apply_fft_reduction(self): def fft_reduction(weights, features):
return single_dim_weights = np.hstack([w.flatten() for w in weights])
x = np.fft.fft(single_dim_weights, n=features)[None, ...]
return x
def __init__(self, width, depth, features, mode='fft', **kwargs): @staticmethod
def random_reduction(weights, features):
x = np.random.rand(features)[None, ...]
return x
def __init__(self, width, depth, features, reduction, **kwargs):
super().__init__(**kwargs) super().__init__(**kwargs)
self.width = width self.width = width
self.depth = depth self.depth = depth
self.features = features self.features = features
self.reduction = reduction
self.compile_params = dict(loss='mse', optimizer='sgd') self.compile_params = dict(loss='mse', optimizer='sgd')
self.apply_reduction = self._apply_fft_reduction if mode.lower()=='fft' else self._apply_mean_reduction
self.model = Sequential()
self.model.add(Dense(units=self.width, input_dim=self.features, **self.keras_params)) self.model.add(Dense(units=self.width, input_dim=self.features, **self.keras_params))
for _ in range(self.depth-1): for _ in range(self.depth-1):
self.model.add(Dense(units=self.width, **self.keras_params)) self.model.add(Dense(units=self.width, **self.keras_params))
@ -355,15 +366,12 @@ class LearningNeuralNetwork(NeuralNetwork):
postfix=["Loss", dict(value=0)]) as bar: postfix=["Loss", dict(value=0)]) as bar:
for epoch in range(epochs): for epoch in range(epochs):
old_weights = self.get_weights() old_weights = self.get_weights()
single_dim_weights = np.hstack([w.flatten() for w in old_weights]) x = self.reduction(old_weights)
x = np.fft.fft(single_dim_weights, n=self.features) history = self.model.fit(x=x, y=x, verbose=0, batch_size=batchsize)
history = self.model.fit(x=x, y=x) bar.postfix[1]["value"] = history.history['loss'][-1]
bar.postfix[1]["value"] = history
bar.update() bar.update()
if __name__ == '__main__': if __name__ == '__main__':
if False: if False:
with FixpointExperiment() as exp: with FixpointExperiment() as exp:
@ -378,8 +386,8 @@ if __name__ == '__main__':
if True: if True:
with IdentLearningExperiment() as exp: with IdentLearningExperiment() as exp:
net = LearningNeuralNetwork(width=2, depth=2, features=2) \ net = LearningNeuralNetwork(width=2, depth=2, features=2, reduction=LearningNeuralNetwork.random_reduction)\
.with_keras_params(activation='linear') \ .with_keras_params(activation='linear') \
.with_params(print_all_weight_updates=False) .with_params(print_all_weight_updates=False)
net.learn(100) net.learn(1000)