def test_eval_points_scaling(Simulator, sample, radius, seed, rng):
eval_points = UniformHypersphere()
if sample:
eval_points = eval_points.sample(500, 3, rng=rng)
model = nengo.Network(seed=seed)
with model:
a = nengo.Ensemble(1, 3, eval_points=eval_points, radius=radius)
with Simulator(model) as sim:
dists = npext.norm(sim.data[a].eval_points, axis=1)
assert np.all(dists <= radius)
assert np.any(dists >= 0.9 * radius)
def test_eval_points_scaling(Simulator, sample, radius, seed, rng, scale):
eval_points = UniformHypersphere()
if sample:
eval_points = eval_points.sample(500, 3, rng=rng)
model = nengo.Network(seed=seed)
with model:
a = nengo.Ensemble(1, 3, radius=radius)
b = nengo.Ensemble(1, 3)
con = nengo.Connection(a, b, eval_points=eval_points,
scale_eval_points=scale)
sim = Simulator(model)
dists = npext.norm(sim.data[con].eval_points, axis=1)
limit = radius if scale else 1.0
assert np.all(dists <= limit)
assert np.any(dists >= 0.9 * limit)
def test_distributionparam_sample_shape():
"""sample_shape dictates the shape of the sample that can be set."""
class Test(object):
dp = params.DistributionParam(default=None, sample_shape=['d1', 10])
d1 = 4
inst = Test()
# Distributions are still cool
inst.dp = UniformHypersphere()
assert isinstance(inst.dp, UniformHypersphere)
# Must be shape (4, 10)
inst.dp = np.ones((4, 10))
assert np.all(inst.dp == np.ones((4, 10)))
with pytest.raises(ValueError):
inst.dp = np.ones((10, 4))
assert np.all(inst.dp == np.ones((4, 10)))
def test_distorarrayparam_sample_shape():
"""sample_shape dictates the shape of the sample that can be set."""
class Test:
dp = DistOrArrayParam("dp", default=None, sample_shape=["d1", 10])
d1 = 4
inst = Test()
# Distributions are still cool
inst.dp = UniformHypersphere()
assert isinstance(inst.dp, UniformHypersphere)
# Must be shape (4, 10)
inst.dp = np.ones((4, 10))
assert np.all(inst.dp == np.ones((4, 10)))
with pytest.raises(ValidationError):
inst.dp = np.ones((10, 4))
assert np.all(inst.dp == np.ones((4, 10)))
def test_distorarrayparam():
"""DistOrArrayParams can be distributions or samples."""
class Test:
dp = DistOrArrayParam("dp", default=None, sample_shape=["*", "*"])
inst = Test()
inst.dp = UniformHypersphere()
assert isinstance(inst.dp, UniformHypersphere)
inst.dp = np.array([[1], [2], [3]])
assert np.all(inst.dp == np.array([[1], [2], [3]]))
with pytest.raises(ValueError):
inst.dp = "a"
# Sample must have correct dims
with pytest.raises(ValueError):
inst.dp = np.array([1])
def random_orthogonal(d, rng=None):
"""Returns a random orthogonal matrix.
Parameters
----------
d : ``integer``
Positive dimension of returned matrix.
rng : :class:`numpy.random.RandomState` or ``None``, optional
Random number generator state.
Returns
-------
samples : ``(d, d) np.array``
Random orthogonal matrix (an orthonormal basis);
linearly transforms any vector into a uniformly sampled
vector on the ``d``--ball with the same L2 norm.
See Also
--------
:class:`.ScatteredHypersphere`
Examples
--------
>>> from nengolib.stats import random_orthogonal, sphere
>>> rng = np.random.RandomState(seed=0)
>>> u = sphere.sample(1000, 3, rng=rng)
>>> u[:, 0] = 0
>>> v = u.dot(random_orthogonal(3, rng=rng))
>>> import matplotlib.pyplot as plt
>>> from mpl_toolkits.mplot3d import Axes3D
>>> ax = plt.subplot(111, projection='3d')
>>> ax.scatter(*u.T, alpha=.5, label="u")
>>> ax.scatter(*v.T, alpha=.5, label="v")
>>> ax.patch.set_facecolor('white')
>>> ax.set_xlim3d(-1, 1)
>>> ax.set_ylim3d(-1, 1)
>>> ax.set_zlim3d(-1, 1)
>>> plt.legend()
>>> plt.show()
"""
rng = np.random if rng is None else rng
m = UniformHypersphere(surface=True).sample(d, d, rng=rng)
u, s, v = svd(m)
return np.dot(u, v)
def test_ndarrays(Simulator, rng, allclose):
encoders = UniformHypersphere(surface=True).sample(10, 1, rng=rng)
max_rates = rng.uniform(200, 400, size=10)
intercepts = rng.uniform(-1, 1, size=10)
eval_points = rng.uniform(-1, 1, size=(100, 1))
with nengo.Network() as net:
ea = nengo.networks.EnsembleArray(
10,
2,
encoders=encoders,
max_rates=max_rates,
intercepts=intercepts,
eval_points=eval_points,
n_eval_points=eval_points.shape[0],
)
with Simulator(net) as sim:
pass
built = sim.data[ea.ea_ensembles[0]]
assert allclose(built.encoders, encoders)
assert allclose(built.max_rates, max_rates)
assert allclose(built.intercepts, intercepts)
assert allclose(built.eval_points, eval_points)
assert built.eval_points.shape == eval_points.shape
with nengo.Network() as net:
# Incorrect shapes don't fail until build
ea = nengo.networks.EnsembleArray(
10,
2,
encoders=encoders,
max_rates=max_rates,
intercepts=intercepts,
eval_points=eval_points[:10],
n_eval_points=eval_points.shape[0],
)
with pytest.raises(ValidationError):
with Simulator(net) as sim:
pass
def test_target_function(Simulator, nl_nodirect, plt, dimensions, radius, seed,
rng):
eval_points = UniformHypersphere().sample(1000, dimensions, rng=rng)
eval_points *= radius
f = lambda x: x**2
targets = f(eval_points)
model = nengo.Network(seed=seed)
with model:
model.config[nengo.Ensemble].neuron_type = nl_nodirect()
inp = nengo.Node(lambda t: np.sin(t * 2 * np.pi) * radius)
ens1 = nengo.Ensemble(40 * dimensions, dimensions, radius=radius)
n1 = nengo.Node(size_in=dimensions)
n2 = nengo.Node(size_in=dimensions)
transform = np.linspace(1, -1, num=dimensions).reshape(-1, 1)
nengo.Connection(inp, ens1, transform=transform)
# pass in eval_points and targets
nengo.Connection(ens1, n1, **target_function(eval_points, targets))
# same, but let the builder apply f
nengo.Connection(ens1, n2, function=f)
probe1 = nengo.Probe(n1, synapse=0.03)
probe2 = nengo.Probe(n2, synapse=0.03)
sim = Simulator(model)
sim.run(0.5)
plt.subplot(2, 1, 1)
plt.plot(sim.trange(), sim.data[probe1])
plt.title('Square manually with target_function')
plt.subplot(2, 1, 2)
plt.plot(sim.trange(), sim.data[probe2])
plt.title('Square by passing in function to connection')
plt.saveas = ('utils.test_connection.test_target_function_%d_%g.pdf' %
(dimensions, radius))
assert np.allclose(sim.data[probe1], sim.data[probe2], atol=0.2 * radius)
def get_eval_points(n_points, dims, rng=None, sort=False):
points = UniformHypersphere(surface=False).sample(n_points, dims, rng=rng)
return points[np.argsort(points[:, 0])] if sort else points
def get_encoders(n_neurons, dims, rng=None):
return UniformHypersphere(surface=True).sample(n_neurons, dims, rng=rng).T
# objective = 'rms'
# objective = 'sp'
if objective == 'nll':
cost = nll_cost_on_inds
error = class_error_on_inds
dout = n_labels
dtp_cheat = True
elif objective == 'rms':
cost = squared_cost_on_inds
error = rms_error_on_inds
dout = n_labels
dtp_cheat = True
elif objective == 'sp':
dout = 50
pointers = UniformHypersphere(surface=True).sample(n_labels,
d=dout,
rng=rng)
dtp_cheat = False
def cost(y, yinds, pointers=pointers):
return pointer_squared_cost_on_inds(y, yinds, pointers)
def error(y, yinds, pointers=pointers):
return pointer_class_error_on_inds(y, yinds, pointers)
din = trainX.shape[1]
sizes = [din] + dhids + [dout]
batch_fn = make_flat_batch_fn(trainX, trainY, n_per_batch)
def test_hypersphere_warns():
with pytest.warns(UserWarning, match="min_magnitude ignored because surface"):
UniformHypersphere(surface=True, min_magnitude=0.1)
def test_hypersphere_errors():
with pytest.raises(ValidationError, match="Must be of type 'bool'"):
UniformHypersphere(0)
with pytest.raises(ValidationError, match="Dimensions must be a positive integer"):
UniformHypersphere().sample(1, d=-1)
def __init__(self, x, ystar, initial_weights, t0=None, t1=None,
n_output=None, o_kind='ensemble',
o_encoders=UniformHypersphere(surface=True),
o_intercepts=Uniform(-1, 0.9), o_rates=Uniform(100, 120),
n_error=None, e_kind='ensemble',
e_encoders=UniformHypersphere(surface=True),
e_intercepts=Uniform(-1, 0.9), e_rates=Uniform(100, 120),
psynapse=None, pdt=None, **kwargs):
super(FeedforwardNetwork, self).__init__(**kwargs)
din = initial_weights[0].shape[0]
dout = initial_weights[-1].shape[1]
dhids = [w.shape[1] for w in initial_weights[:-1]]
self.n_output = n_output
self.n_error = n_error
self.e_kind = e_kind
self.e_encoders = e_encoders
self.e_intercepts = e_intercepts
self.e_rates = e_rates
self.pargs = dict(synapse=psynapse, sample_every=pdt)
with self:
# hidden layers
self.layers = [
nengo.Ensemble(dhid, dhid,
gain=Choice([1.]), bias=Choice([1.]),
label='layer%d' % i)
for i, dhid in enumerate(dhids)]
# output layer
if n_output is None:
self.output = EAIO(
nengo.Node, size_in=dout, label='output')
elif o_kind == 'ensemble':
self.output = EAIO(
nengo.Ensemble, n_output*dout, dout, label='output',
encoders=o_encoders, intercepts=o_intercepts, max_rates=o_rates)
elif o_kind == 'array':
self.output = nengo.networks.EnsembleArray(
n_output, dout, label='output',
encoders=o_encoders, intercepts=o_intercepts, max_rates=o_rates)
self.yp = nengo.Probe(self.output.output, **self.pargs)
# error layer
tmask = ((lambda t: (t > t0) & (t < t1)) if t0 is not None and t1 is not None else \
(lambda t: (t > t0)) if t0 is not None else \
(lambda t: (t < t1)) if t1 is not None else \
(lambda t: (t >= 0)))
nmask = lambda t: 20 * (tmask(t) - 1)
if n_error is None:
ferror = lambda t, x: tmask(t) * x
self.error = EAIO(
nengo.Node, ferror, size_in=dout, label='error')
elif e_kind == 'ensemble':
self.error = EAIO(
nengo.Ensemble, n_error*dout, dout, label='error',
encoders=e_encoders, intercepts=e_intercepts, max_rates=e_rates)
error_switch = nengo.Node(nmask)
nengo.Connection(error_switch, self.error.neuron_input,
transform=np.ones((n_error*dout, 1)), synapse=None)
elif e_kind == 'array':
self.error = nengo.networks.EnsembleArray(
n_error, dout, label='error',
encoders=e_encoders, intercepts=e_intercepts, max_rates=e_rates)
self.error.add_neuron_input()
error_switch = nengo.Node(nmask)
nengo.Connection(error_switch, self.error.neuron_input,
transform=np.ones((n_error*dout, 1)), synapse=None)
else:
raise ValueError(e_kind)
self.ep = nengo.Probe(self.error.output, **self.pargs)
# connections
nengo.Connection(ystar, self.error.input, transform=-1)
with self:
nengo.Connection(self.output.output, self.error.input)
self.conns = []
self.conns.append(nengo.Connection(
x.neurons, self.layers[0].neurons, transform=initial_weights[0].T))
with self:
layers = self.layers
# layers = self.layers + [self.output]
for layer0, layer1, w in zip(layers, layers[1:], initial_weights[1:]):
self.conns.append(nengo.Connection(
layer0.neurons, layer1.neurons, transform=w.T))
self.conns.append(nengo.Connection(
self.layers[-1].neurons, self.output.input, transform=initial_weights[-1].T))
def random_orthogonal(d, rng=None):
rng = np.random if rng is None else rng
m = UniformHypersphere(surface=True).sample(d, d, rng=rng)
u, s, v = svd(m)
return np.dot(u, v)
def test_distributions():
check_init_args(PDF, ["x", "p"])
check_repr(PDF([1, 2, 3], [0.1, 0.8, 0.1]))
assert (repr(PDF(
[1, 2], [0.4, 0.6])) == "PDF(x=array([1., 2.]), p=array([0.4, 0.6]))")
check_init_args(Uniform, ["low", "high", "integer"])
check_repr(Uniform(1, 3))
check_repr(Uniform(1, 4, integer=True))
assert repr(Uniform(0, 1)) == "Uniform(low=0, high=1)"
assert repr(Uniform(
0, 5, integer=True)) == "Uniform(low=0, high=5, integer=True)"
check_init_args(Gaussian, ["mean", "std"])
check_repr(Gaussian(0, 2))
assert repr(Gaussian(1, 0.1)) == "Gaussian(mean=1, std=0.1)"
check_init_args(Exponential, ["scale", "shift", "high"])
check_repr(Exponential(2.0))
check_repr(Exponential(2.0, shift=0.1))
check_repr(Exponential(2.0, shift=0.1, high=10.0))
assert repr(Exponential(2.0)) == "Exponential(scale=2.0)"
check_init_args(UniformHypersphere, ["surface", "min_magnitude"])
check_repr(UniformHypersphere())
check_repr(UniformHypersphere(surface=True))
check_repr(UniformHypersphere(min_magnitude=0.3))
assert repr(UniformHypersphere()) == "UniformHypersphere()"
assert repr(
UniformHypersphere(surface=True)) == "UniformHypersphere(surface=True)"
check_init_args(ScatteredHypersphere,
["surface", "min_magnitude", "base", "method"])
check_repr(ScatteredHypersphere())
check_repr(ScatteredHypersphere(surface=True))
check_repr(ScatteredHypersphere(min_magnitude=0.3))
check_repr(ScatteredHypersphere(base=Uniform(0, 1)))
check_repr(ScatteredHypersphere(method="tfww"))
assert repr(ScatteredHypersphere()) == "ScatteredHypersphere()"
assert (repr(ScatteredHypersphere(
surface=True)) == "ScatteredHypersphere(surface=True)")
assert (repr(ScatteredHypersphere(base=Uniform(0, 1), method="tfww")) ==
"ScatteredHypersphere(base=Uniform(low=0, high=1), method='tfww')")
check_init_args(Choice, ["options", "weights"])
check_repr(Choice([3, 2, 1]))
check_repr(Choice([3, 2, 1], weights=[0.1, 0.2, 0.7]))
assert repr(Choice([1, 2, 3])) == "Choice(options=array([1., 2., 3.]))"
assert (repr(
Choice([1, 2, 3], weights=[0.1, 0.5, 0.4])
) == "Choice(options=array([1., 2., 3.]), weights=array([0.1, 0.5, 0.4]))")
check_init_args(Samples, ["samples"])
check_repr(Samples([3, 2, 1]))
assert repr(Samples([3, 2, 1])) == "Samples(samples=array([3., 2., 1.]))"
check_init_args(SqrtBeta, ["n", "m"])
check_repr(SqrtBeta(3))
check_repr(SqrtBeta(3, m=2))
assert repr(SqrtBeta(3)) == "SqrtBeta(n=3)"
assert repr(SqrtBeta(3, 2)) == "SqrtBeta(n=3, m=2)"
check_init_args(SubvectorLength, ["dimensions", "subdimensions"])
check_repr(SubvectorLength(6))
check_repr(SubvectorLength(6, 2))
assert repr(SubvectorLength(3)) == "SubvectorLength(dimensions=3)"
check_init_args(CosineSimilarity, ["dimensions"])
check_repr(CosineSimilarity(6))
assert repr(CosineSimilarity(6)) == "CosineSimilarity(dimensions=6)"
def test_voja_encoders(Simulator, nl_nodirect, rng, seed, allclose):
"""Tests that voja changes active encoders to the input."""
n = 200
learned_vector = np.asarray([0.3, -0.4, 0.6])
learned_vector /= np.linalg.norm(learned_vector)
n_change = n // 2 # modify first half of the encoders
# Set the first half to always fire with random encoders, and the
# remainder to never fire due to their encoder's dot product with the input
intercepts = np.asarray([-1] * n_change + [0.99] * (n - n_change))
rand_encoders = UniformHypersphere(surface=True).sample(
n_change, len(learned_vector), rng=rng)
encoders = np.append(rand_encoders, [-learned_vector] * (n - n_change),
axis=0)
m = nengo.Network(seed=seed)
with m:
m.config[nengo.Ensemble].neuron_type = nl_nodirect()
u = nengo.Node(output=learned_vector)
x = nengo.Ensemble(
n,
dimensions=len(learned_vector),
intercepts=intercepts,
encoders=encoders,
max_rates=nengo.dists.Uniform(300.0, 400.0),
radius=2.0,
) # to test encoder scaling
conn = nengo.Connection(u,
x,
synapse=None,
learning_rule_type=Voja(learning_rate=1e-1))
p_enc = nengo.Probe(conn.learning_rule, "scaled_encoders")
p_enc_ens = nengo.Probe(x, "scaled_encoders")
with Simulator(m) as sim:
sim.run(1.0)
t = sim.trange()
tend = t > 0.5
# Voja's rule relies on knowing exactly how the encoders were scaled
# during the build process, because it modifies the scaled_encoders signal
# proportional to this factor. Therefore, we should check that its
# assumption actually holds.
encoder_scale = (sim.data[x].gain / x.radius)[:, np.newaxis]
assert allclose(sim.data[x].encoders,
sim.data[x].scaled_encoders / encoder_scale)
# Check that the last half kept the same encoders throughout the simulation
assert allclose(sim.data[p_enc][0, n_change:], sim.data[p_enc][:,
n_change:])
# and that they are also equal to their originally assigned value
assert allclose(sim.data[p_enc][0, n_change:] / encoder_scale[n_change:],
-learned_vector)
# Check that the first half converged to the input
assert allclose(
sim.data[p_enc][tend, :n_change] / encoder_scale[:n_change],
learned_vector,
atol=0.01,
)
# Check that encoders probed from ensemble equal encoders probed from Voja
assert allclose(sim.data[p_enc], sim.data[p_enc_ens])
def test_hypersphere_dimension_fail(rng):
with pytest.raises(ValueError):
UniformHypersphere(0).sample(1, 0)
class Ensemble(NengoObject):
"""A group of neurons that collectively represent a vector.
Parameters
----------
n_neurons : int
The number of neurons.
dimensions : int
The number of representational dimensions.
radius : int, optional (Default: 1.0)
The representational radius of the ensemble.
encoders : Distribution or (n_neurons, dimensions) array_like, optional \
(Default: UniformHypersphere(surface=True))
The encoders used to transform from representational space
to neuron space. Each row is a neuron's encoder; each column is a
representational dimension.
intercepts : Distribution or (n_neurons,) array_like, optional \
(Default: ``nengo.dists.Uniform(-1.0, 1.0)``)
The point along each neuron's encoder where its activity is zero. If
``e`` is the neuron's encoder, then the activity will be zero when
``dot(x, e) <= c``, where ``c`` is the given intercept.
max_rates : Distribution or (n_neurons,) array_like, optional \
(Default: ``nengo.dists.Uniform(200, 400)``)
The activity of each neuron when the input signal ``x`` is magnitude 1
and aligned with that neuron's encoder ``e``;
i.e., when ``dot(x, e) = 1``.
eval_points : Distribution or (n_eval_points, dims) array_like, optional \
(Default: ``nengo.dists.UniformHypersphere()``)
The evaluation points used for decoder solving, spanning the interval
(-radius, radius) in each dimension, or a distribution from which
to choose evaluation points.
n_eval_points : int, optional (Default: None)
The number of evaluation points to be drawn from the ``eval_points``
distribution. If None, then a heuristic is used to determine
the number of evaluation points.
neuron_type : `~nengo.neurons.NeuronType`, optional \
(Default: ``nengo.LIF()``)
The model that simulates all neurons in the ensemble
(see `~nengo.neurons.NeuronType`).
gain : Distribution or (n_neurons,) array_like (Default: None)
The gains associated with each neuron in the ensemble. If None, then
the gain will be solved for using ``max_rates`` and ``intercepts``.
bias : Distribution or (n_neurons,) array_like (Default: None)
The biases associated with each neuron in the ensemble. If None, then
the gain will be solved for using ``max_rates`` and ``intercepts``.
noise : Process, optional (Default: None)
Random noise injected directly into each neuron in the ensemble
as current. A sample is drawn for each individual neuron on
every simulation step.
normalize_encoders : bool, optional (Default: True)
Indicates whether the encoders should be normalized.
label : str, optional (Default: None)
A name for the ensemble. Used for debugging and visualization.
seed : int, optional (Default: None)
The seed used for random number generation.
Attributes
----------
bias : Distribution or (n_neurons,) array_like or None
The biases associated with each neuron in the ensemble.
dimensions : int
The number of representational dimensions.
encoders : Distribution or (n_neurons, dimensions) array_like
The encoders, used to transform from representational space
to neuron space. Each row is a neuron's encoder, each column is a
representational dimension.
eval_points : Distribution or (n_eval_points, dims) array_like
The evaluation points used for decoder solving, spanning the interval
(-radius, radius) in each dimension, or a distribution from which
to choose evaluation points.
gain : Distribution or (n_neurons,) array_like or None
The gains associated with each neuron in the ensemble.
intercepts : Distribution or (n_neurons) array_like or None
The point along each neuron's encoder where its activity is zero. If
``e`` is the neuron's encoder, then the activity will be zero when
``dot(x, e) <= c``, where ``c`` is the given intercept.
label : str or None
A name for the ensemble. Used for debugging and visualization.
max_rates : Distribution or (n_neurons,) array_like or None
The activity of each neuron when ``dot(x, e) = 1``,
where ``e`` is the neuron's encoder.
n_eval_points : int or None
The number of evaluation points to be drawn from the ``eval_points``
distribution. If None, then a heuristic is used to determine
the number of evaluation points.
n_neurons : int or None
The number of neurons.
neuron_type : NeuronType
The model that simulates all neurons in the ensemble
(see ``nengo.neurons``).
noise : Process or None
Random noise injected directly into each neuron in the ensemble
as current. A sample is drawn for each individual neuron on
every simulation step.
radius : int
The representational radius of the ensemble.
seed : int or None
The seed used for random number generation.
"""
probeable = ('decoded_output', 'input', 'scaled_encoders')
n_neurons = IntParam('n_neurons', low=1)
dimensions = IntParam('dimensions', low=1)
radius = NumberParam('radius', default=1.0, low=1e-10)
encoders = DistOrArrayParam('encoders',
default=UniformHypersphere(surface=True),
sample_shape=('n_neurons', 'dimensions'))
intercepts = DistOrArrayParam('intercepts',
default=Uniform(-1.0, 1.0),
optional=True,
sample_shape=('n_neurons', ))
max_rates = DistOrArrayParam('max_rates',
default=Uniform(200, 400),
optional=True,
sample_shape=('n_neurons', ))
eval_points = DistOrArrayParam('eval_points',
default=UniformHypersphere(),
sample_shape=('*', 'dimensions'))
n_eval_points = IntParam('n_eval_points', default=None, optional=True)
neuron_type = NeuronTypeParam('neuron_type', default=LIF())
gain = DistOrArrayParam('gain',
default=None,
optional=True,
sample_shape=('n_neurons', ))
bias = DistOrArrayParam('bias',
default=None,
optional=True,
sample_shape=('n_neurons', ))
noise = ProcessParam('noise', default=None, optional=True)
normalize_encoders = BoolParam('normalize_encoders',
default=True,
optional=True)
def __init__(self,
n_neurons,
dimensions,
radius=Default,
encoders=Default,
intercepts=Default,
max_rates=Default,
eval_points=Default,
n_eval_points=Default,
neuron_type=Default,
gain=Default,
bias=Default,
noise=Default,
normalize_encoders=Default,
label=Default,
seed=Default):
super(Ensemble, self).__init__(label=label, seed=seed)
self.n_neurons = n_neurons
self.dimensions = dimensions
self.radius = radius
self.encoders = encoders
self.intercepts = intercepts
self.max_rates = max_rates
self.n_eval_points = n_eval_points
self.eval_points = eval_points
self.bias = bias
self.gain = gain
self.neuron_type = neuron_type
self.noise = noise
self.normalize_encoders = normalize_encoders
def __getitem__(self, key):
return ObjView(self, key)
def __len__(self):
return self.dimensions
@property
def neurons(self):
"""A direct interface to the neurons in the ensemble."""
return Neurons(self)
@neurons.setter
def neurons(self, dummy):
raise ReadonlyError(attr="neurons", obj=self)
@property
def size_in(self):
"""The dimensionality of the ensemble."""
return self.dimensions
@property
def size_out(self):
"""The dimensionality of the ensemble."""
return self.dimensions
def test_hypersphere_warns(rng):
with pytest.warns(UserWarning):
UniformHypersphere(surface=True, min_magnitude=0.1)
class Ensemble(NengoObject):
"""A group of neurons that collectively represent a vector.
Parameters
----------
n_neurons : int
The number of neurons.
dimensions : int
The number of representational dimensions.
radius : int, optional
The representational radius of the ensemble.
encoders : Distribution or ndarray (`n_neurons`, `dimensions`), optional
The encoders, used to transform from representational space
to neuron space. Each row is a neuron's encoder, each column is a
representational dimension.
intercepts : Distribution or ndarray (`n_neurons`), optional
The point along each neuron's encoder where its activity is zero. If
e is the neuron's encoder, then the activity will be zero when
dot(x, e) <= c, where c is the given intercept.
max_rates : Distribution or ndarray (`n_neurons`), optional
The activity of each neuron when dot(x, e) = 1, where e is the neuron's
encoder.
eval_points : Distribution or ndarray (`n_eval_points`, `dims`), optional
The evaluation points used for decoder solving, spanning the interval
(-radius, radius) in each dimension, or a distribution from which to
choose evaluation points. Default: ``UniformHypersphere``.
n_eval_points : int, optional
The number of evaluation points to be drawn from the `eval_points`
distribution. If None (the default), then a heuristic is used to
determine the number of evaluation points.
neuron_type : Neurons, optional
The model that simulates all neurons in the ensemble.
noise : StochasticProcess, optional
Random noise injected directly into each neuron in the ensemble
as current. A sample is drawn for each individual neuron on
every simulation step.
seed : int, optional
The seed used for random number generation.
label : str, optional
A name for the ensemble. Used for debugging and visualization.
"""
n_neurons = IntParam(default=None, low=1)
dimensions = IntParam(default=None, low=1)
radius = NumberParam(default=1, low=1e-10)
neuron_type = NeuronTypeParam(default=LIF())
encoders = DistributionParam(default=UniformHypersphere(surface=True),
sample_shape=('n_neurons', 'dimensions'))
intercepts = DistributionParam(default=Uniform(-1.0, 1.0),
optional=True,
sample_shape=('n_neurons', ))
max_rates = DistributionParam(default=Uniform(200, 400),
optional=True,
sample_shape=('n_neurons', ))
n_eval_points = IntParam(default=None, optional=True)
eval_points = DistributionParam(default=UniformHypersphere(),
sample_shape=('*', 'dimensions'))
bias = DistributionParam(default=None,
optional=True,
sample_shape=('n_neurons', ))
gain = DistributionParam(default=None,
optional=True,
sample_shape=('n_neurons', ))
noise = StochasticProcessParam(default=None, optional=True)
seed = IntParam(default=None, optional=True)
label = StringParam(default=None, optional=True)
def __init__(self,
n_neurons,
dimensions,
radius=Default,
encoders=Default,
intercepts=Default,
max_rates=Default,
eval_points=Default,
n_eval_points=Default,
neuron_type=Default,
gain=Default,
bias=Default,
noise=Default,
seed=Default,
label=Default):
self.n_neurons = n_neurons
self.dimensions = dimensions
self.radius = radius
self.encoders = encoders
self.intercepts = intercepts
self.max_rates = max_rates
self.label = label
self.n_eval_points = n_eval_points
self.eval_points = eval_points
self.bias = bias
self.gain = gain
self.neuron_type = neuron_type
self.noise = noise
self.seed = seed
self._neurons = Neurons(self)
def __getitem__(self, key):
return ObjView(self, key)
def __len__(self):
return self.dimensions
@property
def neurons(self):
return self._neurons
@neurons.setter
def neurons(self, dummy):
raise AttributeError("neurons cannot be overwritten.")
@property
def probeable(self):
return ["decoded_output", "input"]
@property
def size_in(self):
return self.dimensions
@property
def size_out(self):
return self.dimensions
def test_scattered_hypersphere(dims, surface, seed, plt):
scipy_special = pytest.importorskip("scipy.special")
n = 3000
dists = [
UniformHypersphere(surface=surface),
ScatteredHypersphere(surface=surface, method="sct"),
ScatteredHypersphere(surface=surface, method="sct-approx"),
ScatteredHypersphere(surface=surface, method="tfww"),
]
assert isinstance(dists[0], UniformHypersphere)
xx = [] # generated points, for each dist
times = [] # time taken to generate the points, for each dist
for dist in dists:
rng = np.random.RandomState(seed)
timer = time.time()
x = dist.sample(n, d=dims, rng=rng)
timer = time.time() - timer
rng.shuffle(x) # shuffle so we can compute distances in blocks without bias
xx.append(x)
times.append(timer)
dd = [] # distance to the nearest point for each point, for each dist
rr = [] # radii (norms) of all the generated points, for each dist
for x in xx:
# compute distances in blocks for efficiency (this means we're not actually
# getting the minimum distance, just a proxy)
n_split = 1000
d_min = []
for i in range(0, n, n_split):
xi = x[i : i + n_split]
d2 = ((xi[:, :, None] - xi.T[None, :, :]) ** 2).sum(axis=1)
np.fill_diagonal(d2, np.inf)
d_min.append(np.sqrt(d2.min(axis=1)))
d_min = np.concatenate(d_min)
dd.append(d_min)
rr.append(np.sqrt((x ** 2).sum(axis=1)))
# compute the approximate distance between points if they were evenly spread
volume = np.pi ** (0.5 * dims) / scipy_special.gamma(0.5 * dims + 1)
if surface:
volume *= dims
even_distance = (volume / n) ** (1 / (dims - 1 if surface else dims))
# --- plots
colors = ["b", "g", "r", "m", "c"]
plt.subplot(211)
bins = np.linspace(np.min(dd), np.max(dd), 31)
for i, d in enumerate(dd):
histogram, _ = np.histogram(d, bins=bins)
plt.plot(
0.5 * (bins[:-1] + bins[1:]),
histogram,
colors[i],
)
plt.plot([d.min()], [0], colors[i] + "x")
plt.plot([even_distance], [0], "kx")
plt.title("surface=%s, dims=%d, n=%d" % (surface, dims, n))
plt.subplot(212)
bins = np.linspace(0, 1.1, 31)
for i, r in enumerate(rr):
histogram, _ = np.histogram(r, bins=bins)
plt.plot(
0.5 * (bins[:-1] + bins[1:]),
histogram,
colors[i],
label=f"{dists[i]}: t={times[i]:0.2e}",
)
plt.legend()
# --- checks
uniform_min = dd[0].min()
for i, dist in enumerate(dists):
if i == 0:
continue
# check that we're significantly better than UniformHypersphere
d_min = dd[i].min()
assert d_min > 1.2 * uniform_min, str(dist)
# check that all surface points are on the surface
if surface:
assert np.allclose(rr[i], 1.0, atol=1e-5), str(dist)
