def test_connect_additional_input(num_neurons, expected_output):
neurons = neuron.Neurons(num_neurons)
other = neuron.Neurons(num_neurons)
neurons.set('feedback', other)
other.output.values = expected_output
assert all(neurons.get('feedback').values == expected_output)
def test_connect_input(num_neurons, expected_output):
neurons = neuron.Neurons(num_neurons)
other = neuron.Neurons(num_neurons)
neurons.input = other
other.output.values = expected_output
assert all(neurons.input.values == expected_output)
def test_standard_computation():
compute = computation.StandardComputation(threshold=1.2)
n1 = neuron.Neurons(3)
n2 = neuron.Neurons(2, computation=compute)
synapses = np.array([[0.8, 0.4], [0.6, 0.8], [0.5, 0.4]])
n2.set(neuron.Neurons.MAIN_INPUT,
n1,
synapse_function=lambda inp_shape, out_shape: synapses)
n1.output.values = np.array([1, 0, 1])
output = n2.compute()
expected = [True, False]
assert all(output == expected)
def test_sparse_computation_with_fixed_number():
compute = computation.SparseComputation(n=0.25)
n1 = neuron.Neurons(3)
n2 = neuron.Neurons(4, computation=compute)
synapses = np.array([[0.8, 0.4, 0.5, 0.4], [0.6, 0.8, 0.9, 0.2],
[0.5, 0.4, 0.5, 0.7]])
n2.set(neuron.Neurons.MAIN_INPUT,
n1,
synapse_function=lambda inp_shape, out_shape: synapses)
n1.output.values = np.array([1, 0, 1])
output = n2.compute()
expected = [True, False, False, False]
assert all(output == expected)
def test_hebbian_learning():
hl = learning.HebbianLearning(increase=0.1, decrease=0.05)
n1 = neuron.Neurons(3)
n2 = neuron.Neurons(2, learning=hl)
synapses = np.array([[0.8, 0.4], [0.6, 0.8], [0.5, 0.4]])
n2.set(neuron.Neurons.MAIN_INPUT,
n1,
synapse_function=lambda inp_shape, out_shape: synapses)
n1.output.values = np.array([1, 0, 1])
n2.output.values = np.array([1, 0])
n2.learn()
expected_synapses = np.array([[0.90, 0.35], [0.55, 0.80], [0.60, 0.35]])
testing.assert_allclose(n2.input.synapses, expected_synapses)
def create_brain(architecture,
sensor_dim,
computation=None,
learning=None,
random_seed=0):
np.random.seed(random_seed)
cortex = neuron.FeedForward([neuron.Neurons(i) for i in architecture])
return brain.Brain(cortex,
create_sensor(sensor_dim),
computation=computation,
learning=learning)
def decode_neurons(neurons_proto):
"""Obtain an instance of Neurons from a protobuf.
Args:
neurons_proto: The protobuf instance representing the neurons
Returns:
The neurons decoded from the protobuf
"""
layers = [neuron.Neurons(n) for n in neurons_proto.layer]
return neuron.FeedForward(layers)
def neurons():
return neuron.Neurons(4, None)
def layers():
return neuron.FeedForward([neuron.Neurons(i, None) for i in (4, 4, 4, 4)])
def test_loop_back():
layers = [neuron.Neurons(i) for i in [10, 10, 10]]
neuron.LoopBack(layers, 'loopback')
for layer in layers:
layer.get('loopback')._connected_output == layer.output
def test_feed_backward():
layers = [neuron.Neurons(i) for i in [10, 10, 10]]
neuron.FeedBackward(layers, 'feedback')
for i, layer in enumerate(layers[:-1]):
layer.get('feedback')._connected_output == layers[i + 1].output
def test_feed_forward_custom_input():
layers = [neuron.Neurons(i) for i in [10, 10, 10]]
neuron.FeedForward(layers, 'extra')
for i, layer in enumerate(layers[:-1]):
layers[i + 1].get('extra')._connected_output == layer.output
def test_feed_forward():
layers = [neuron.Neurons(i) for i in [10, 10, 10]]
neuron.FeedForward(layers)
for i, layer in enumerate(layers[:-1]):
layers[i + 1].input._connected_output == layer.output
def test_no_connections():
layers = [neuron.Neurons(i) for i in [10, 10, 10]]
for layer in layers:
with pytest.raises(IndexError):
layer.input.values
def test_null_inputs(num_neurons):
neurons = neuron.Neurons(num_neurons)
with pytest.raises(IndexError):
neurons.input.values