How to use the nengo.Ensemble function in nengo
To help you get started, we’ve selected a few nengo examples, based on popular ways it is used in public projects.
Secure your code as it's written. Use Snyk Code to scan source code in minutes - no build needed - and fix issues immediately.
studywolf / blog / Nengo scripting / Nengo 2 / arm_control / nengo_arm.py View on Github 
<svg viewBox="0 0 100 100" height="100%" width="100%">
<line style="stroke:black" y2="{y2}" x2="{x2}" y1="{y1}" x1="{x1}"></line>
<line style="stroke:black" y2="{y3}" x2="{x3}" y1="{y2}" x1="{x2}"></line>
<circle fill="black" stroke-width="1" stroke="black" r="1.5" cy="{y3}" cx="{x3}"></circle>
</svg>
'''.format(**locals())
# end of visualization code ---------------------------------------------
return data
arm_node = nengo.Node(output=arm_func, size_in=2)
# specify torque input to arm
input_node = nengo.Node(output=[1, .1])
# to send a target to an ensemble which then connections to the arm
ens = nengo.Ensemble(n_neurons=500, dimensions=2, radius=20)
nengo.Connection(input_node, ens[:2]) # to send target info to ensemble
# connect ens to arm
nengo.Connection(ens, arm_node)#, function=some_function)
# --------------------------------------------------------
if run_in_GUI:
# to run in GUI, comment out next 4 lines for running without GUI
import nengo_gui
nengo_gui.GUI(model=model, filename=__file__, locals=locals(),
interactive=False, allow_file_change=False).start()
import sys
sys.exit()
else:
# to run in command line
with model:copter = Multicopter()
copter.start()
q_target = nengo.Node(lambda t: q_target_value, label='q_target')
q = nengo.Node(lambda t: q_value, label='q')
dq = nengo.Node(lambda t: dq_value, label='dq')
dq_target = nengo.Node(None, size_in=1, label='dq_target')
nengo.Connection(q_target, dq_target, synapse=None, transform=1) # direct feed-forward synapse
nengo.Connection(q_target, dq_target, synapse=0, transform=-1) # minimal time-step delay
q_err = nengo.Ensemble(n_neurons=N_NEURONS, dimensions=1, label='q_err')
nengo.Connection(q_target, q_err, synapse=None)
nengo.Connection(q, q_err, synapse=None, transform=-1)
q_err_integral = nengo.Ensemble(n_neurons=N_NEURONS, dimensions=1, radius=INTEGRAL_RADIUS,
label='q_err_integral')
nengo.Connection(q_err, q_err_integral, synapse=INTEGRAL_SYNAPSE, transform=INTEGRAL_SYNAPSE)
nengo.Connection(q_err_integral, q_err_integral, synapse=INTEGRAL_SYNAPSE, transform=1)
x = nengo.Ensemble(n_neurons=1200, dimensions=3, radius=30)
u = nengo.Node(None, size_in=1, label='u') # output
nengo.Connection(q_err, u, transform=KP, synapse=None)
nengo.Connection(q_err_integral, u, transform=KI, synapse=None)
dq_err = nengo.Ensemble(n_neurons=N_NEURONS, dimensions=1, label='dq_err')
nengo.Connection(dq_target, dq_err, synapse=None)
nengo.Connection(dq, dq_err, synapse=None, transform=-1)with model:
q_target = nengo.Node(lambda t: self.q_target_value, label='q_target')
q = nengo.Node(lambda t: self.q_value, label='q')
dq = nengo.Node(lambda t: self.dq_value, label='dq')
dq_target = nengo.Node(None, size_in=self.n_dims, label='dq_target')
nengo.Connection(q_target, dq_target, synapse=None, transform=1) # direct feed-forward synapse
nengo.Connection(q_target, dq_target, synapse=0, transform=-1) # minimal time-step delay
q_err = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=self.n_dims, label='q_err')
nengo.Connection(q_target, q_err, synapse=None)
nengo.Connection(q, q_err, synapse=None, transform=-1)
q_err_integral = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=self.n_dims, radius=self.integral_radius,
label='q_err_integral')
nengo.Connection(q_err, q_err_integral, synapse=self.integral_synapse, transform=self.integral_synapse)
nengo.Connection(q_err_integral, q_err_integral, synapse=self.integral_synapse, transform=1)
u = nengo.Node(None, size_in=self.n_dims, label='u') # output
nengo.Connection(q_err, u, transform=self.Kp, synapse=None)
nengo.Connection(q_err_integral, u, transform=self.Ki, synapse=None)
dq_err = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=self.n_dims, label='dq_err')
nengo.Connection(dq_target, dq_err, synapse=None)
nengo.Connection(dq, dq_err, synapse=None, transform=-1)
nengo.Connection(dq_err, u, transform=self.Kd, synapse=None)zeroidxs = np.where(rate==0)[0] # indices of neurons that don't fire at all
## build an ensemble exactly as in the test file simulation
## and find the bises of those neurons
## NOTE: Set the seeds and other params manually below,
## as they could be ambiguous from name of the file
import nengo
Nexc, N, reprRadius, nrngain = 3000, 2, 5, 2
seedR0, seedR2 = 2, 4
gain_bias_set = True
#biaslow, biashigh = 1 - nrngain, 1 + nrngain
biaslow, biashigh = -nrngain, nrngain
print('building model')
mainModel = nengo.Network(label="Single layer network", seed=seedR0)
with mainModel:
ratorOut = nengo.Ensemble( Nexc, dimensions=N, radius=reprRadius,
neuron_type=nengo.neurons.LIF(),
bias=nengo.dists.Uniform(biaslow,biashigh), gain=np.ones(Nexc)*nrngain,
#max_rates=nengo.dists.Uniform(200, 400),
noise=None, seed=seedR2, label='ratorOut' )
sim = nengo.Simulator(mainModel,dt)
biases = sim.data[ratorOut].bias
zerofiringbiases = biases[zeroidxs]
gains = sim.data[ratorOut].gain
zerofiringgains = gains[zeroidxs]
if gain_bias_set: histrange, biasrange = 5, 5
else: histrange, biasrange = 500, 100
fig = plt.figure(facecolor='w')
ax1 = plt.subplot(231)
vals,_,_ = ax1.hist(gains,bins=50,range=(0,histrange),color='k',histtype='step')
ax1.set_xlabel('all gains')with self.model:
q_target = nengo.Node(lambda t: self.q_target_value, label='q_target')
q = nengo.Node(lambda t: self.q_value, label='q')
dq = nengo.Node(lambda t: self.dq_value, label='dq')
dq_target = nengo.Node(None, size_in=1, label='dq_target')
nengo.Connection(q_target, dq_target, synapse=None, transform=1) # direct feed-forward synapse
nengo.Connection(q_target, dq_target, synapse=0, transform=-1) # minimal time-step delay
q_err = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=1, label='q_err')
nengo.Connection(q_target, q_err, synapse=None)
nengo.Connection(q, q_err, synapse=None, transform=-1)
q_err_integral = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=1, radius=integral_radius,
label='q_err_integral')
nengo.Connection(q_err, q_err_integral, synapse=integral_synapse, transform=integral_synapse)
nengo.Connection(q_err_integral, q_err_integral, synapse=integral_synapse, transform=1)
u = nengo.Node(None, size_in=1, label='u') # output
nengo.Connection(q_err, u, transform=self.Kp, synapse=None)
nengo.Connection(q_err_integral, u, transform=Ki, synapse=None)
dq_err = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=1, label='dq_err')
nengo.Connection(dq_target, dq_err, synapse=None)
nengo.Connection(dq, dq_err, synapse=None, transform=-1)
nengo.Connection(dq_err, u, transform=self.Kd, synapse=None)seed=seedR2)
else:
processIn = None
processOut = None
nodeIn = nengo.Node( size_in=N, output = lambda timeval,currval: inpfn(timeval) )
# with zero bias, at reprRadius, if you want 50Hz, gain=1.685, if 100Hz, gain=3.033, if 400Hz, 40.5
#nrngain = 1.5#2#3.033
# input layer from which feedforward weights to ratorOut are computed
ratorIn = nengo.Ensemble( Nexc, dimensions=N, radius=reprRadiusIn,
neuron_type=nengo.neurons.LIF(),
#bias=nengo.dists.Uniform(-nrngain,nrngain), gain=np.ones(Nexc)*nrngain,
max_rates=nengo.dists.Uniform(200, 400),
noise=processIn, seed=seedR1, label='ratorIn' )
nengo.Connection(nodeIn, ratorIn, synapse=None) # No filtering here as no filtering/delay in the plant/arm
# another layer with learning incorporated
ratorOut = nengo.Ensemble( Nexc, dimensions=N, radius=reprRadius,
neuron_type=nengo.neurons.LIF(),
#bias=nengo.dists.Uniform(-nrngain,nrngain), gain=np.ones(Nexc)*nrngain,
max_rates=nengo.dists.Uniform(200, 400),
noise=processOut, seed=seedR2, label='ratorOut' )
if trialClamp:
# clamp ratorIn and ratorOut at the end of each trial (Tperiod) for 100ms.
# Error clamped below during end of the trial for 100ms.
clampValsZeros = np.zeros(Nexc)
clampValsNegs = -100.*np.ones(Nexc)
endTrialClamp = nengo.Node(lambda t: clampValsZeros if (t%Tperiod)<(Tperiod-Tclamp) else clampValsNegs)
nengo.Connection(endTrialClamp,ratorIn.neurons,synapse=1e-3)
nengo.Connection(endTrialClamp,ratorOut.neurons,synapse=1e-3)
# fast synapse for fast-reacting clamp
if plastDecoders:start = 2*i
end = ((2*(i+1))-1)+1
# Connected the world's move sensor to pacman's move sensor
nengo.Connection(sensor[start:end], pacPlayer.move)
# Created Food ensemble
food = nengo.Ensemble(n_neurons = 100, dimensions = 2)
nengo.Connection(pacPlayer.detect_food, food, synapse = 0.)
# Connected pacman's food node to the world's ensemble
nengo.Connection(food[0], sensor[end-1], transform = 2, synapse = 0.)
nengo.Connection(food[1], sensor[start], transform = 3, synapse = 0.)
# Created obstacles ensemble
obstacles = nengo.Ensemble(n_neurons = 50, dimensions = 3, radius = 4)
nengo.Connection(pacPlayer.obstacles[[1,2,3]], obstacles, transform = 0.5, synapse = 0.)
# turn away from walls
def avoid(x):
return 1*(x[2] - x[0])
# Connected the obstacles node to the world's ensemble
nengo.Connection(obstacles, sensor[end-1], function=avoid, synapse = 0.)
# avoid crashing into walls
def ahead(x):
return 1*(x[1] - 0.5)
# Connected the obstacles node to the world's ensemble
nengo.Connection(obstacles, sensor[start], function=ahead, synapse = 0.)
enemy = nengo.Ensemble(n_neurons = 50, dimensions = 2)
nengo.Connection(pacPlayer.detect_enemy, enemy, synapse = 0.)# sense food
# - angle is the direction to the food
# - radius is the strength (1.0/distance)
food = nengo.Ensemble(n_neurons=100, dimensions=2)
nengo.Connection(pacnet.detect_food, food, synapse = 0.)
# turn towards food
nengo.Connection(food[0], move[1], transform=2)
# move towards food
nengo.Connection(food[1], move[0], transform=3)
# sense obstacles
# - distance to obstacles on left, front-left, front, front-right, and right
# - maximum distance is 4
obstacles = nengo.Ensemble(n_neurons=50, dimensions=3, radius=4)
nengo.Connection(pacnet.obstacles[[1,2,3]], obstacles, transform=0.5, synapse = 0.)
# turn away from walls
def avoid(x):
return 1*(x[2] - x[0])
nengo.Connection(obstacles, move[1], function=avoid)
# avoid crashing into walls
def ahead(x):
return 1*(x[1] - 0.5)
nengo.Connection(obstacles, move[0], function=ahead)
# detect enemies
enemy = nengo.Ensemble(n_neurons=50, dimensions=2)
nengo.Connection(pacnet.detect_enemy, enemy, synapse = 0.)processIn = nengo.processes.WhiteNoise(
default_size_out=Nexc, default_dt = dt,
dist=nengo.dists.Gaussian(0 if (dynNoiseMean is None) else dynNoiseMean, dynNoise),
seed=seedR1)
processOut = nengo.processes.WhiteNoise(
default_size_out=Nexc, default_dt = dt,
dist=nengo.dists.Gaussian(0 if (dynNoiseMean is None) else dynNoiseMean, dynNoise),
seed=seedR2)
else:
processIn = None
processOut = None
nodeIn = nengo.Node( size_in=N, output = lambda timeval,currval: inpfn(timeval) )
# with zero bias, at reprRadius, if you want 50Hz, gain=1.685, if 100Hz, gain=3.033, if 400Hz, 40.5
#nrngain = 1.5#2#3.033
# input layer from which feedforward weights to ratorOut are computed
ratorIn = nengo.Ensemble( Nexc, dimensions=N, radius=reprRadiusIn,
neuron_type=nengo.neurons.LIF(),
#bias=nengo.dists.Uniform(-nrngain,nrngain), gain=np.ones(Nexc)*nrngain,
max_rates=nengo.dists.Uniform(200, 400),
noise=processIn, seed=seedR1, label='ratorIn' )
nengo.Connection(nodeIn, ratorIn, synapse=None) # No filtering here as no filtering/delay in the plant/arm
# another layer with learning incorporated
ratorOut = nengo.Ensemble( Nexc, dimensions=N, radius=reprRadius,
neuron_type=nengo.neurons.LIF(),
#bias=nengo.dists.Uniform(-nrngain,nrngain), gain=np.ones(Nexc)*nrngain,
max_rates=nengo.dists.Uniform(200, 400),
noise=processOut, seed=seedR2, label='ratorOut' )
if trialClamp:
# clamp ratorIn and ratorOut at the end of each trial (Tperiod) for 100ms.
# Error clamped below during end of the trial for 100ms.
clampValsZeros = np.zeros(Nexc)self.n_neurons = n_neurons
self.integral_synapse = integral_synapse
self.integral_radius = integral_radius
self.seed = seed
with model:
q_target = nengo.Node(lambda t: self.q_target_value, label='q_target')
q = nengo.Node(lambda t: self.q_value, label='q')
dq = nengo.Node(lambda t: self.dq_value, label='dq')
dq_target = nengo.Node(None, size_in=self.n_dims, label='dq_target')
nengo.Connection(q_target, dq_target, synapse=None, transform=1) # direct feed-forward synapse
nengo.Connection(q_target, dq_target, synapse=0, transform=-1) # minimal time-step delay
q_err = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=self.n_dims, label='q_err')
nengo.Connection(q_target, q_err, synapse=None)
nengo.Connection(q, q_err, synapse=None, transform=-1)
q_err_integral = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=self.n_dims, radius=self.integral_radius,
label='q_err_integral')
nengo.Connection(q_err, q_err_integral, synapse=self.integral_synapse, transform=self.integral_synapse)
nengo.Connection(q_err_integral, q_err_integral, synapse=self.integral_synapse, transform=1)
u = nengo.Node(None, size_in=self.n_dims, label='u') # output
nengo.Connection(q_err, u, transform=self.Kp, synapse=None)
nengo.Connection(q_err_integral, u, transform=self.Ki, synapse=None)
dq_err = nengo.Ensemble(n_neurons=self.n_neurons, dimensions=self.n_dims, label='dq_err')
nengo.Connection(dq_target, dq_err, synapse=None)nengo
Tools for building and simulating large-scale neural models
Package Health Score
58 / 100Popular nengo functions
- nengo.base
- nengo.builder.connection.get_eval_points
- nengo.builder.connection.get_targets
- nengo.Connection
- nengo.decoders
- nengo.dists
- nengo.Ensemble
- nengo.LIF
- nengo.Lowpass
- nengo.Network
- nengo.networks
- nengo.networks.EnsembleArray
- nengo.neurons
- nengo.neurons.LIF
- nengo.Node
- nengo.PES
- nengo.Probe
- nengo.Simulator
- nengo.utils.distributions.Uniform