def test_weights_built(self):
"""Test a build using a weights-based solver."""
# Create the network
with nengo.Network():
a = nengo.Ensemble(200, 2)
b = nengo.Ensemble(400, 2)
a_b = nengo.Connection(a,
b,
solver=nengo.solvers.Lstsq(weights=True))
# Create the model and built the pre-synaptic Ensemble
model = builder.Model()
model.rng = np.random
model.seeds[a] = 1
model.seeds[b] = 2
model.seeds[a_b] = 3
ensemble.build_ensemble(model, a)
ensemble.build_ensemble(model, b)
# Now build the connection and check that the params seem sensible
ensemble.build_from_ensemble_connection(model, a_b)
# Check that the params stored in the model are correct
params = model.params[a_b]
assert params.decoders.shape == (200, 400)
def test_to_global_inhibition(self):
"""Test that the transmission parameters are modified for a global
inhibition connection.
"""
# Create the network
with nengo.Network():
a = nengo.Ensemble(200, 3)
b = nengo.Ensemble(300, 1)
a_b = nengo.Connection(a,
b.neurons,
transform=[[1.0, 0.5, 0.2]] * b.n_neurons)
# Create the model and built the pre-synaptic Ensemble
model = builder.Model()
model.rng = np.random
model.seeds[a] = 1
model.seeds[a_b] = 2
ensemble.build_ensemble(model, a)
# Now build the connection and check that the params seem sensible
tparams = ensemble.build_from_ensemble_connection(model, a_b)
assert tparams.decoders.shape == (1, 200)
# Check that the params stored in the model are correct
params = model.params[a_b]
assert np.all(params.transform == a_b.transform)
def test_build_ensemble_lif(self, n_neurons, size_in):
"""Test building LIF ensembles."""
# Create a Nengo ensemble to build
ens = nengo.Ensemble(n_neurons, size_in, add_to_container=False)
# Create a model
model = builder.Model()
model.seeds[ens] = 1
# Build the ensemble
ensemble.build_ensemble(model, ens)
# Check that the built ensemble was inserted into the params and that
# the parameters are (loosely) as expected.
assert model.params[ens].eval_points is not None
assert (model.params[ens].encoders.shape ==
model.params[ens].scaled_encoders.shape ==
(n_neurons, size_in))
assert (model.params[ens].intercepts.shape ==
model.params[ens].max_rates.shape ==
model.params[ens].gain.shape == model.params[ens].bias.shape ==
(n_neurons, ))
# Check that a new object was inserted into the objects dictionary
assert isinstance(model.object_operators[ens], operators.EnsembleLIF)
def test_arbitrary_neuron_sink(self, source):
"""Test that standard connections to neurons return an appropriate
sink.
We have no plan to support arbitrary connections to neurons, but we
allow them at this stage because they may later become global
inhibition connections when we optimise out passthrough Nodes.
"""
with nengo.Network():
a = nengo.Ensemble(100, 2)
b = nengo.Ensemble(100, 4)
if source == "neurons":
a_b = nengo.Connection(a.neurons,
b.neurons,
transform=np.eye(100))
else:
a_b = nengo.Connection(a,
b.neurons,
transform=[[1.0, 0.5]] * 99 +
[[0.5, 1.0]])
# Create a model with the Ensemble for b in it
model = builder.Model()
b_ens = operators.EnsembleLIF(b)
model.object_operators[b] = b_ens
# Get the sink, check that an appropriate target is return
sink = ensemble.get_neurons_sink(model, a_b)
assert sink.target.obj is b_ens
assert sink.target.port is ensemble.EnsembleInputPort.neurons
def test_standard_build(self):
"""Test relatively standard build."""
# Create the network
with nengo.Network():
a = nengo.Ensemble(200, 3)
b = nengo.Node(lambda t, x: None, size_in=2)
a_b = nengo.Connection(a[:2],
b,
transform=np.array([[0.5, 0], [0.0, 0.0]]))
# Create the model and built the pre-synaptic Ensemble
model = builder.Model()
model.rng = np.random
model.seeds[a] = 1
model.seeds[a_b] = 2
ensemble.build_ensemble(model, a)
# Now build the connection and check that the params seem sensible
tparams = ensemble.build_from_ensemble_connection(model, a_b)
assert tparams.decoders.shape == (2, 200)
assert np.all(tparams.decoders[1, :] == 0.0)
# Check that the params stored in the model are correct
params = model.params[a_b]
assert params.decoders.shape == (200, 2)
assert np.all(params.transform == a_b.transform)
assert np.all(params.eval_points == model.params[a].eval_points)
assert params.solver_info is not None
def test_probe_input(self):
"""Test probing the input of an Ensemble."""
with nengo.Network():
a = nengo.Ensemble(100, 3)
p = nengo.Probe(a, "input")
# Create an empty model to build the probe into
model = builder.Model()
model.rng = np.random
model.seeds[p] = 1
# Build the probe
ensemble.build_ensemble_probe(model, p)
def test_probe_spike_slice(self):
with nengo.Network() as net:
a = nengo.Ensemble(300, 1)
p = nengo.Probe(a.neurons[:100], "spikes")
# Create an empty model to build the probe into
model = builder.Model()
model.build(net)
# Assert that we added the probe to the list of local probes and
# nothing else
assert model.object_operators[a].local_probes == [p]
assert len(model.object_operators) == 1
assert list(model.connection_map.get_signals()) == []
def test_probe_output_no_sampling(self):
"""Test that probing the output of an Ensemble generates a new
connection and a new object.
"""
with nengo.Network() as net:
a = nengo.Ensemble(100, 3)
p = nengo.Probe(a)
# Create an empty model to build the probe into
model = builder.Model()
model.build(net)
# Check that a new object was added to the model
vs = model.object_operators[p]
assert vs.sample_every == 1
def test_only_bias(self):
"""Build an ensemble with only bias specified."""
# Create a Nengo ensemble to build
ens = nengo.Ensemble(1, 1, add_to_container=False)
ens.bias = np.array([-0.5])
# Create a model
model = builder.Model()
model.seeds[ens] = 1
# Build the ensemble
ensemble.build_ensemble(model, ens)
# Check that parameters are (loosely) as expected.
assert model.params[ens].bias == ens.bias # pragma : no cover
def test_normal_source(self):
# Create a network and standard model
with nengo.Network():
a = nengo.Ensemble(100, 2)
b = nengo.Ensemble(200, 4)
a_b = nengo.Connection(a, b[1:3])
# Create a model with the Ensemble for b in it
model = builder.Model()
a_ens = operators.EnsembleLIF(a)
model.object_operators[a] = a_ens
source = ensemble.get_ensemble_source(model, a_b)
assert source.target.obj is a_ens
assert source.target.port is OutputPort.standard
def test_encoder_learnt_sink(self):
# Create a network and standard model
with nengo.Network():
a = nengo.Ensemble(100, 2)
b = nengo.Ensemble(100, 2)
a_b = nengo.Connection(a, b)
a_b.learning_rule_type = nengo.Voja()
# Create a model with the Ensemble for b in it
model = builder.Model()
b_ens = operators.EnsembleLIF(b)
model.object_operators[b] = b_ens
sink = ensemble.get_ensemble_sink(model, a_b)
assert sink.target.obj is b_ens
assert sink.target.port is ensemble.EnsembleInputPort.learnt
def test_neurons_source():
"""Test that neurons sources are sane."""
with nengo.Network():
a = nengo.Ensemble(100, 2)
b = nengo.Ensemble(100, 4)
a_b = nengo.Connection(a.neurons, b.neurons, transform=np.eye(100))
# Create a model with the Ensemble for a in it
model = builder.Model()
a_ens = operators.EnsembleLIF(a)
model.object_operators[a] = a_ens
# Get the source, check that an appropriate target is return
source = ensemble.get_neurons_source(model, a_b)
assert source.target.obj is a_ens
assert source.target.port is ensemble.EnsembleOutputPort.neurons
def test_refractory_time(self):
"""Check that probing refractory time modifies the local_probes list on
the operator, but does nothing else.
"""
with nengo.Network() as net:
a = nengo.Ensemble(300, 1)
p = nengo.Probe(a.neurons, "refractory_time")
# Create an empty model to build the probe into
model = builder.Model()
model.build(net)
# Assert that we added the probe to the list of local probes and
# nothing else
assert model.object_operators[a].local_probes == [p]
assert len(model.object_operators) == 1
assert len(model.connections_signals) == 0
def test_probe_voja_scaled_encoders(self):
# Create a network and standard model
with nengo.Network() as net:
a = nengo.Ensemble(100, 2)
b = nengo.Ensemble(100, 2)
a_b = nengo.Connection(a, b)
a_b.learning_rule_type = nengo.Voja()
p = nengo.Probe(a_b.learning_rule, "scaled_encoders")
# Create an empty model to build the probe into
model = builder.Model()
model.build(net)
# Assert that we added the probe to the list of local probes and
# nothing else
assert model.object_operators[b].local_probes == [p]
assert len(model.object_operators) == 2
def test_normal_sink_for_process_node(self):
"""Test that sinks for most connections into Ensembles do nothing
special.
"""
# Create a network and standard model
with nengo.Network():
a = nengo.Node(nengo.processes.WhiteNoise(), size_out=4)
b = nengo.Ensemble(200, 4)
a_b = nengo.Connection(a, b)
# Create a model with the Ensemble for b in it
model = builder.Model()
b_ens = operators.EnsembleLIF(b)
model.object_operators[b] = b_ens
# Get the sink, check that an appropriate target is return
sink = ensemble.get_ensemble_sink(model, a_b)
assert sink.target.obj is b_ens
assert sink.target.port is InputPort.standard
def test_constant_node_sink_with_slice(self):
"""Test that connections from constant valued Nodes to Ensembles are
optimised out correctly.
"""
# Create a network and standard model
with nengo.Network():
a = nengo.Node([0.5, 1.0])
b = nengo.Ensemble(200, 2)
a_b = nengo.Connection(a[0], b[1])
# Create a model with the Ensemble for b in it
model = builder.Model()
b_ens = operators.EnsembleLIF(b)
model.object_operators[b] = b_ens
# Check that no sink is created but that the direct input is modified
assert np.all(b_ens.direct_input == np.zeros(2))
assert ensemble.get_ensemble_sink(model, a_b) is None
assert np.all(b_ens.direct_input == [0.0, 0.5])
def test_decoder_learnt_source(self):
# Create a network and standard model
with nengo.Network():
a = nengo.Ensemble(100, 2)
b = nengo.Ensemble(100, 2)
e = nengo.Ensemble(100, 2)
a_b = nengo.Connection(a, b)
a_b.learning_rule_type = nengo.PES()
e_l = nengo.Connection(e, a_b.learning_rule)
# Create a model with the Ensemble for b in it
model = builder.Model()
a_ens = operators.EnsembleLIF(a)
model.object_operators[a] = a_ens
source = ensemble.get_ensemble_source(model, a_b)
assert source.target.obj is a_ens
assert source.target.port is ensemble.EnsembleOutputPort.learnt
def test_with_encoders_and_gain_bias(self):
"""Test that the encoders we provide are used (albeit scaled)"""
# Create a Nengo ensemble to build
ens = nengo.Ensemble(1, 1, add_to_container=False)
ens.radius = 10.0
ens.encoders = np.array([[1.0]])
ens.gain = np.array([0.5])
ens.bias = np.array([0.0])
# Create a model
model = builder.Model()
model.seeds[ens] = 1
# Build the ensemble
ensemble.build_ensemble(model, ens)
# Check that parameters are (loosely) as expected.
assert model.params[ens].encoders == ens.encoders
assert model.params[ens].gain == ens.gain
assert model.params[ens].bias == ens.bias
assert model.params[ens].scaled_encoders == ens.encoders * (0.5 / 10)
def test_probe_output_with_sampling(self, with_slice):
"""Test that probing the output of an Ensemble generates a new
connection and a new object.
"""
with nengo.Network() as net:
a = nengo.Ensemble(100, 3)
if not with_slice:
p = nengo.Probe(a, sample_every=0.0023)
else:
p = nengo.Probe(a[0:1], sample_every=0.0023)
# Create an empty model to build the probe into
model = builder.Model()
model.build(net)
# Check that a new connection was added and built
assert len(list(model.connection_map.get_signals())) == 1
# Check that a new object was added to the model
vs = model.object_operators[p]
assert isinstance(vs, operators.ValueSink)
assert vs.probe is p
def test_encoder_learning_rule_sink(self):
"""Test that sinks for most connections into Ensembles do nothing
special.
"""
# Create a network and standard model
with nengo.Network():
a = nengo.Ensemble(100, 2)
b = nengo.Ensemble(100, 2)
e = nengo.Ensemble(100, 1)
a_b = nengo.Connection(a, b)
a_b.learning_rule_type = nengo.Voja()
e_l = nengo.Connection(e, a_b.learning_rule)
# Create a model with the Ensemble for b in it
model = builder.Model()
b_ens = operators.EnsembleLIF(b)
model.object_operators[b] = b_ens
# Get the sink, check that an appropriate target is return
sink = ensemble.get_learning_rule_sink(model, e_l)
assert sink.target.obj is b_ens
assert sink.target.port is a_b.learning_rule
def test_global_inhibition_sink(self):
"""Test that sinks are correctly determined for connections which are
global inhibition connections.
"""
# Create a model with a global inhibition connection
with nengo.Network():
a = nengo.Ensemble(100, 2)
b = nengo.Ensemble(200, 4)
a_b = nengo.Connection(a, b.neurons, transform=[[1.0, 0.5]] * 200)
# Create a model with the Ensemble for b in it
model = builder.Model()
b_ens = operators.EnsembleLIF(b)
model.object_operators[b] = b_ens
decs = mock.Mock()
evals = mock.Mock()
si = mock.Mock()
# Get the sink, check that an appropriate target is return
sink = ensemble.get_neurons_sink(model, a_b)
assert sink.target.obj is b_ens
assert sink.target.port is ensemble.EnsembleInputPort.global_inhibition
