def __init__(self,
n_input,
n_output,
is_cuda,
n_neurons_per_layer=3,
n_hidden_layers=2):
super(ProbabilisticFeedForwardAlternative, self).__init__()
self.n_input = n_input
self.n_output = n_output
self.n_neurons_per_layer = n_neurons_per_layer
self.n_hidden_layers = n_hidden_layers
self.is_cuda = is_cuda
self.activation = get_activation()
self.n_layers = n_hidden_layers
in_features = n_input
# hidden layers
for i in range(n_hidden_layers, 0, -1):
layer = Linear(in_features=in_features,
out_features=n_neurons_per_layer,
is_cuda=is_cuda)
setattr(self, f'layer_{i}', layer)
setattr(self, f'activation_{i}', self.activation)
in_features = n_neurons_per_layer
layer = Linear(in_features=in_features,
out_features=n_output,
is_cuda=is_cuda)
setattr(self, f'layer_0', layer)
def __init__(self, genome: Genome):
super(StochasticNetworkOld, self).__init__()
self.n_output = genome.n_output
self.n_input = genome.n_input
self.nodes = genome.node_genes
self.connections = genome.connection_genes
self.activation = get_activation()
layers_dict = _transform_genome_to_layers(nodes=self.nodes,
connections=self.connections,
n_output=self.n_output)
self.n_layers = len(layers_dict)
self._set_network_layers(layers=layers_dict)
def __init__(self, genome: dict, n_input, n_output, activation_type):
super(ComplexStochasticNetworkJupyneat, self).__init__()
self.n_output = n_output
self.n_input = n_input
self.nodes = genome['nodes']
self.connections = genome['connections']
self.activation = get_activation(activation=activation_type)
self.layers = transform_genome_to_layers(genome=genome,
n_input=n_input,
n_output=n_output)
self.n_layers = len(self.layers)
self._set_network_layers(layers=self.layers)
self._cache = {}
def __init__(self, n_input, n_output, n_neurons_per_layer=3, n_hidden_layers=2):
super(FeedForward, self).__init__()
self.n_input = n_input
self.n_output = n_output
self.n_neurons_per_layer = n_neurons_per_layer
self.n_hidden_layers = n_hidden_layers
self.activation = get_activation()
self.n_layers = n_hidden_layers
in_features = n_input
# hidden layers
for i in range(n_hidden_layers, 0, -1):
layer = nn.Linear(in_features=in_features, out_features=n_neurons_per_layer)
setattr(self, f'layer_{i}', layer)
setattr(self, f'activation_{i}', self.activation)
in_features = n_neurons_per_layer
# output layer
layer = nn.Linear(in_features=in_features, out_features=n_output)
setattr(self, f'layer_0', layer)
def __init__(self, genome: Genome, is_trainable=False, is_cuda=False):
super(ComplexStochasticNetwork, self).__init__()
self.is_trainable = is_trainable
self.is_cuda = is_cuda
self.genome = genome
self.n_output = genome.n_output
self.n_input = genome.n_input
self.nodes = genome.node_genes
self.connections = genome.connection_genes
self.config = genome.genome_config
self.activation = get_activation(
activation=self.config.node_activation)
self.layers = transform_genome_to_layers(genome=genome)
self.layers_masks = generate_layer_masks(self.layers,
genome,
fix_std=self.config.fix_std)
self.n_layers = len(self.layers)
self._set_network_layers(layers=self.layers,
layers_masks=self.layers_masks)
self._cache = {}