def test_context():
with Network():
# make some arbitrary object for the reservoirs. it's okay that it's
# not within the same network, because we won't simulate it
a = nengo.Node(size_in=1)
with Network() as outer:
with Network() as inner:
assert Reservoir(a, a).network is inner
assert Reservoir(a, a, network=inner).network is inner
assert Reservoir(a, a, network=outer).network is outer
assert Reservoir(a, a).network is outer
assert Reservoir(a, a, network=inner).network is inner
assert Reservoir(a, a, network=outer).network is outer
with pytest.raises(NetworkContextError):
Reservoir(a, a)
assert Reservoir(a, a, network=inner).network is inner
assert Reservoir(a, a, network=outer).network is outer
def test_bad_inputs_outputs():
with Network():
only_out = nengo.Node(output=[0])
no_out = nengo.Node(size_out=0)
okay = nengo.Ensemble(1, 1)
bad = nengo.Network()
with pytest.raises(ValueError): # must contain at least one input
Reservoir([], okay)
with pytest.raises(ValueError): # must contain at least one output
Reservoir(okay, [])
with pytest.raises(ValueError): # must contain at least one input
Reservoir(only_out, okay)
with pytest.raises(ValueError): # must contain at least one output
Reservoir(okay, no_out)
with pytest.raises(TypeError): # must be connectable object
Reservoir(bad, okay)
with pytest.raises(TypeError): # must be connectable object
Reservoir(okay, bad)
def test_multiple(Simulator, seed, plt):
# check that multiple dimensions go to multiple input nodes
train_t = 2.0
test_t = 0.5
dt = 0.001
process = WhiteSignal(max(train_t, test_t), high=10, rms=0.3)
w = [1, -1, 0.6, -0.3]
function = (lambda x: w[0]*x[:, 0] + w[1]*x[:, 1] + w[2]*x[:, 2] +
w[3]*x[:, 1]*x[:, 2])
with Network() as model:
stim = nengo.Node(output=process, size_out=3)
# isolate the objects within the reservoir so that we can train
# it even after connecting a node into it
with Network():
a = nengo.Node(size_in=1)
b = nengo.Node(size_in=2)
c = nengo.Node(size_in=2, output=lambda t, x: x[0]*x[1])
nengo.Connection(b, c, synapse=None)
p_a = nengo.Probe(a, synapse=None)
p_b = nengo.Probe(b, synapse=None)
p_c = nengo.Probe(c, synapse=None)
res = Reservoir([a, b], [a, b, c])
nengo.Connection(stim[0], a, synapse=None)
nengo.Connection(stim[1:], b, synapse=None)
p = nengo.Probe(res.output, synapse=None)
p_stim = nengo.Probe(stim, synapse=None)
d, info = res.train(
function, train_t, dt, process, seed=seed,
solver=nengo.solvers.LstsqL2(reg=1e-6))
assert np.allclose(np.squeeze(d), w)
assert res.size_in == 3
assert res.size_mid == 4
assert res.size_out == 1
sim, data_in, data_mid = info['sim'], info['data_in'], info['data_mid']
assert np.allclose(sim.data[p_a], data_in[:, 0:1])
assert np.allclose(sim.data[p_b], data_in[:, 1:])
assert np.allclose(sim.data[p_c], data_in[:, 1:2] * data_in[:, 2:])
assert np.allclose(
np.hstack((sim.data[p_a], sim.data[p_b], sim.data[p_c])), data_mid)
with Simulator(model, dt=dt, seed=seed+1) as sim:
sim.run(test_t)
ideal = function(sim.data[p_stim])
plt.figure()
plt.plot(sim.trange(), sim.data[p_a], label="a")
plt.plot(sim.trange(), sim.data[p_b], label="b")
plt.plot(sim.trange(), sim.data[p_c], label="c")
plt.plot(sim.trange(), sim.data[p], label="Output")
plt.plot(sim.trange(), ideal, label="Ideal")
plt.legend()
assert np.allclose(np.squeeze(sim.data[p]), ideal)
def test_bad_function():
with Network():
a = nengo.Node(size_in=1)
res = Reservoir(a, a)
with pytest.raises(RuntimeError): # expected signal length 1000
res.train(lambda x: 0, 1.0, 0.001, WhiteSignal(1.0, high=10))
def test_seed_warning():
with Network():
ens = nengo.Ensemble(100, 1)
res = Reservoir(ens, ens)
with warns(UserWarning):
res.train(lambda x: x, 1.0, 0.001, WhiteSignal(1.0, high=10))
def test_basic(Simulator, seed, plt):
# solve for a standard nengo connection using a feed-forward reservoir
train_t = 5.0
test_t = 0.5
dt = 0.001
n_neurons = 100
synapse = 0.01
process = WhiteSignal(max(train_t, test_t), high=10, rms=0.3)
def function(x): return x**2
with Network() as model:
ens = nengo.Ensemble(n_neurons, 1, seed=seed) # <- must have seed!
res = Reservoir(ens, ens.neurons, synapse)
# Solve for the readout that approximates a function of the *filtered*
# stimulus. We include a lowpass here because the final RMSE will be
# with respect to the lowpass stimulus, which is also consistent
# with what the NEF is doing. But in a general recurrent reservoir
# this filter could hypothetically be anything.
res.train(
lambda x: function(Lowpass(synapse).filt(x, dt=dt)),
train_t, dt, process, seed=seed+1)
assert res.size_in == 1
assert res.size_mid == n_neurons
assert res.size_out == 1
# Validation
_, (_, _, check_output) = res.run(
test_t, dt, process, seed=seed+2)
stim = nengo.Node(output=process)
output = nengo.Node(size_in=1)
nengo.Connection(stim, ens, synapse=None)
nengo.Connection(ens, output, function=function, synapse=None)
# note the reservoir output already includes a synapse
p_res = nengo.Probe(res.output, synapse=None)
p_normal = nengo.Probe(output, synapse=synapse)
p_stim = nengo.Probe(stim, synapse=synapse)
with Simulator(model, dt=dt, seed=seed+2) as sim:
sim.run(test_t)
# Since the seed for the two test processes were the same, the validation
# run should produce the same output as the test simulation.
assert np.allclose(check_output, sim.data[p_res])
ideal = function(sim.data[p_stim])
plt.figure()
plt.plot(sim.trange(), sim.data[p_res], label="Reservoir")
plt.plot(sim.trange(), sim.data[p_normal], label="Standard")
plt.plot(sim.trange(), ideal, label="Ideal")
plt.legend()
assert np.allclose(rmse(sim.data[p_res], ideal),
rmse(sim.data[p_normal], ideal), atol=1e-2)
def test_multiple(Simulator, seed, plt):
# check that multiple dimensions go to multiple input nodes
train_t = 2.0
test_t = 0.5
dt = 0.001
process = WhiteSignal(max(train_t, test_t), high=10, rms=0.3)
w = [1, -1, 0.6, -0.3]
function = (lambda x: w[0] * x[:, 0] + w[1] * x[:, 1] + w[2] * x[:, 2] + w[
3] * x[:, 1] * x[:, 2])
with Network() as model:
stim = nengo.Node(output=process, size_out=3)
# isolate the objects within the reservoir so that we can train
# it even after connecting a node into it
with Network():
a = nengo.Node(size_in=1)
b = nengo.Node(size_in=2)
c = nengo.Node(size_in=2, output=lambda t, x: x[0] * x[1])
nengo.Connection(b, c, synapse=None)
p_a = nengo.Probe(a, synapse=None)
p_b = nengo.Probe(b, synapse=None)
p_c = nengo.Probe(c, synapse=None)
res = Reservoir([a, b], [a, b, c])
nengo.Connection(stim[0], a, synapse=None)
nengo.Connection(stim[1:], b, synapse=None)
p = nengo.Probe(res.output, synapse=None)
p_stim = nengo.Probe(stim, synapse=None)
d, info = res.train(function,
train_t,
dt,
process,
seed=seed,
solver=nengo.solvers.LstsqL2(reg=1e-6))
assert np.allclose(np.squeeze(d), w)
assert res.size_in == 3
assert res.size_mid == 4
assert res.size_out == 1
sim, data_in, data_mid = info['sim'], info['data_in'], info['data_mid']
assert np.allclose(sim.data[p_a], data_in[:, 0:1])
assert np.allclose(sim.data[p_b], data_in[:, 1:])
assert np.allclose(sim.data[p_c], data_in[:, 1:2] * data_in[:, 2:])
assert np.allclose(
np.hstack((sim.data[p_a], sim.data[p_b], sim.data[p_c])), data_mid)
with Simulator(model, dt=dt, seed=seed + 1) as sim:
sim.run(test_t)
ideal = function(sim.data[p_stim])
plt.figure()
plt.plot(sim.trange(), sim.data[p_a], label="a")
plt.plot(sim.trange(), sim.data[p_b], label="b")
plt.plot(sim.trange(), sim.data[p_c], label="c")
plt.plot(sim.trange(), sim.data[p], label="Output")
plt.plot(sim.trange(), ideal, label="Ideal")
plt.legend()
assert np.allclose(np.squeeze(sim.data[p]), ideal)
def test_bad_function():
with Network():
a = nengo.Node(size_in=1)
res = Reservoir(a, a)
with pytest.raises(RuntimeError): # expected signal length 1000
res.train(lambda x: 0, 1.0, 0.001, WhiteSignal(1.0, high=10))
def test_seed_warning():
with Network():
ens = nengo.Ensemble(100, 1)
res = Reservoir(ens, ens)
with warns(UserWarning):
res.train(lambda x: x, 1.0, 0.001, WhiteSignal(1.0, high=10))
def test_basic(Simulator, seed, plt):
# solve for a standard nengo connection using a feed-forward reservoir
train_t = 5.0
test_t = 0.5
dt = 0.001
n_neurons = 100
synapse = 0.01
process = WhiteSignal(max(train_t, test_t), high=10, rms=0.3)
def function(x):
return x**2
with Network() as model:
ens = nengo.Ensemble(n_neurons, 1, seed=seed) # <- must have seed!
res = Reservoir(ens, ens.neurons, synapse)
# Solve for the readout that approximates a function of the *filtered*
# stimulus. We include a lowpass here because the final RMSE will be
# with respect to the lowpass stimulus, which is also consistent
# with what the NEF is doing. But in a general recurrent reservoir
# this filter could hypothetically be anything.
res.train(lambda x: function(Lowpass(synapse).filt(x, dt=dt)),
train_t,
dt,
process,
seed=seed + 1)
assert res.size_in == 1
assert res.size_mid == n_neurons
assert res.size_out == 1
# Validation
_, (_, _, check_output) = res.run(test_t, dt, process, seed=seed + 2)
stim = nengo.Node(output=process)
output = nengo.Node(size_in=1)
nengo.Connection(stim, ens, synapse=None)
nengo.Connection(ens, output, function=function, synapse=None)
# note the reservoir output already includes a synapse
p_res = nengo.Probe(res.output, synapse=None)
p_normal = nengo.Probe(output, synapse=synapse)
p_stim = nengo.Probe(stim, synapse=synapse)
with Simulator(model, dt=dt, seed=seed + 2) as sim:
sim.run(test_t)
# Since the seed for the two test processes were the same, the validation
# run should produce the same output as the test simulation.
assert np.allclose(check_output, sim.data[p_res])
ideal = function(sim.data[p_stim])
plt.figure()
plt.plot(sim.trange(), sim.data[p_res], label="Reservoir")
plt.plot(sim.trange(), sim.data[p_normal], label="Standard")
plt.plot(sim.trange(), ideal, label="Ideal")
plt.legend()
assert np.allclose(rmse(sim.data[p_res], ideal),
rmse(sim.data[p_normal], ideal),
atol=1e-2)
def __init__(self,
n_neurons,
dimensions,
recurrent_synapse=0.005,
readout_synapse=None,
radii=1.0,
gain=1.25,
rng=None,
neuron_type=Tanh(),
include_bias=True,
ens_seed=None,
label=None,
seed=None,
add_to_container=None,
**ens_kwargs):
"""Initializes the Echo State Network.
Parameters
----------
n_neurons : int
The number of neurons to use in the reservoir.
dimensions : int
The dimensionality of the input signal.
recurrent_synapse : nengo.synapses.Synapse (Default: ``0.005``)
Synapse used to filter the recurrent connection.
readout_synapse : nengo.synapses.Synapse (Default: ``None``)
Optional synapse to filter all of the outputs before solving
for the linear readout. This is included in the connection to the
``output`` Node created within the network.
radii : scalar or array_like, optional (Default: ``1``)
The radius of each dimension of the input signal, used to normalize
the incoming connection weights.
gain : scalar, optional (Default: ``1.25``)
A scalar gain on the recurrent connection weight matrix.
rng : ``numpy.random.RandomState``, optional (Default: ``None``)
Random state used to initialize all weights.
neuron_type : ``nengo.neurons.NeuronType`` optional \
(Default: ``Tanh()``)
Neuron model to use within the reservoir.
include_bias : ``bool`` (Default: ``True``)
Whether to include a bias current to the neural nonlinearity.
This should be ``False`` if the neuron model already has a bias,
e.g., ``LIF`` or ``LIFRate``.
ens_seed : int, optional (Default: ``None``)
Seed passed to the ensemble of neurons.
"""
Network.__init__(self, label, seed, add_to_container)
self.n_neurons = n_neurons
self.dimensions = dimensions
self.recurrent_synapse = recurrent_synapse
self.radii = radii # TODO: make array or scalar parameter?
self.gain = gain
self.rng = np.random if rng is None else rng
self.neuron_type = neuron_type
self.include_bias = include_bias
self.W_in = (self.rng.rand(self.n_neurons, self.dimensions) -
0.5) / self.radii
if self.include_bias:
self.W_bias = self.rng.rand(self.n_neurons, 1) - 0.5
else:
self.W_bias = np.zeros((self.n_neurons, 1))
self.W = self.rng.rand(self.n_neurons, self.n_neurons) - 0.5
self.W *= self.gain / max(abs(eig(self.W)[0]))
with self:
self.ensemble = nengo.Ensemble(self.n_neurons,
1,
neuron_type=self.neuron_type,
seed=ens_seed,
**ens_kwargs)
self.input = nengo.Node(size_in=self.dimensions)
pool = self.ensemble.neurons
nengo.Connection(self.input,
pool,
transform=self.W_in,
synapse=None)
nengo.Connection( # note the bias will be active during training
nengo.Node(output=1, label="bias"),
pool,
transform=self.W_bias,
synapse=None)
nengo.Connection(self.ensemble.neurons,
pool,
transform=self.W,
synapse=self.recurrent_synapse)
Reservoir.__init__(self,
self.input,
pool,
readout_synapse=readout_synapse,
network=self)
def __init__(self, n_neurons, dimensions, recurrent_synapse=0.005,
readout_synapse=None, radii=1.0, gain=1.25, rng=None,
neuron_type=Tanh(), include_bias=True, ens_seed=None,
label=None, seed=None, add_to_container=None, **ens_kwargs):
"""Initializes the Echo State Network.
Parameters
----------
n_neurons : int
The number of neurons to use in the reservoir.
dimensions : int
The dimensionality of the input signal.
recurrent_synapse : nengo.synapses.Synapse (Default: ``0.005``)
Synapse used to filter the recurrent connection.
readout_synapse : nengo.synapses.Synapse (Default: ``None``)
Optional synapse to filter all of the outputs before solving
for the linear readout. This is included in the connection to the
``output`` Node created within the network.
radii : scalar or array_like, optional (Default: ``1``)
The radius of each dimension of the input signal, used to normalize
the incoming connection weights.
gain : scalar, optional (Default: ``1.25``)
A scalar gain on the recurrent connection weight matrix.
rng : ``numpy.random.RandomState``, optional (Default: ``None``)
Random state used to initialize all weights.
neuron_type : ``nengo.neurons.NeuronType`` optional \
(Default: ``Tanh()``)
Neuron model to use within the reservoir.
include_bias : ``bool`` (Default: ``True``)
Whether to include a bias current to the neural nonlinearity.
This should be ``False`` if the neuron model already has a bias,
e.g., ``LIF`` or ``LIFRate``.
ens_seed : int, optional (Default: ``None``)
Seed passed to the ensemble of neurons.
"""
Network.__init__(self, label, seed, add_to_container)
self.n_neurons = n_neurons
self.dimensions = dimensions
self.recurrent_synapse = recurrent_synapse
self.radii = radii # TODO: make array or scalar parameter?
self.gain = gain
self.rng = np.random if rng is None else rng
self.neuron_type = neuron_type
self.include_bias = include_bias
self.W_in = (
self.rng.rand(self.n_neurons, self.dimensions) - 0.5) / self.radii
if self.include_bias:
self.W_bias = self.rng.rand(self.n_neurons, 1) - 0.5
else:
self.W_bias = np.zeros((self.n_neurons, 1))
self.W = self.rng.rand(self.n_neurons, self.n_neurons) - 0.5
self.W *= self.gain / max(abs(eig(self.W)[0]))
with self:
self.ensemble = nengo.Ensemble(
self.n_neurons, 1, neuron_type=self.neuron_type, seed=ens_seed,
**ens_kwargs)
self.input = nengo.Node(size_in=self.dimensions)
pool = self.ensemble.neurons
nengo.Connection(
self.input, pool, transform=self.W_in, synapse=None)
nengo.Connection( # note the bias will be active during training
nengo.Node(output=1, label="bias"), pool,
transform=self.W_bias, synapse=None)
nengo.Connection(
self.ensemble.neurons, pool, transform=self.W,
synapse=self.recurrent_synapse)
Reservoir.__init__(
self, self.input, pool, readout_synapse=readout_synapse,
network=self)
