def make(net,name='System',neurons=100,A=[[0]],tau_feedback=0.1):
A=numeric.array(A)
assert len(A.shape)==2
assert A.shape[0]==A.shape[1]
dimensions=A.shape[0]
state=net.make(name,neurons,dimensions)
Ap=A*tau_feedback+numeric.identity(dimensions)
net.connect(state,state,transform=Ap,pstc=tau_feedback)
if net.network.getMetaData("linear") == None:
net.network.setMetaData("linear", HashMap())
linears = net.network.getMetaData("linear")
linear=HashMap(4)
linear.put("name", name)
linear.put("neurons", neurons)
linear.put("A", MU.clone(A))
linear.put("tau_feedback", tau_feedback)
linears.put(name, linear)
if net.network.getMetaData("templates") == None:
net.network.setMetaData("templates", ArrayList())
templates = net.network.getMetaData("templates")
templates.add(name)
if net.network.getMetaData("templateProjections") == None:
net.network.setMetaData("templateProjections", HashMap())
templateproj = net.network.getMetaData("templateProjections")
templateproj.put(name, name)
def make(net, name='System', neurons=100, A=[[0]], tau_feedback=0.1):
A = numeric.array(A)
assert len(A.shape) == 2
assert A.shape[0] == A.shape[1]
dimensions = A.shape[0]
state = net.make(name, neurons, dimensions)
Ap = A * tau_feedback + numeric.identity(dimensions)
net.connect(state, state, transform=Ap, pstc=tau_feedback)
if net.network.getMetaData("linear") == None:
net.network.setMetaData("linear", HashMap())
linears = net.network.getMetaData("linear")
linear = HashMap(4)
linear.put("name", name)
linear.put("neurons", neurons)
linear.put("A", MU.clone(A))
linear.put("tau_feedback", tau_feedback)
linears.put(name, linear)
if net.network.getMetaData("templates") == None:
net.network.setMetaData("templates", ArrayList())
templates = net.network.getMetaData("templates")
templates.add(name)
if net.network.getMetaData("templateProjections") == None:
net.network.setMetaData("templateProjections", HashMap())
templateproj = net.network.getMetaData("templateProjections")
templateproj.put(name, name)
def getValue(self):
return MU.clone(self.matrix)
def make_convolution(
self,
name,
A,
B,
C,
N_per_D,
quick=False,
encoders=[[1, 1], [1, -1], [-1, 1], [-1, -1]],
radius=3,
pstc_out=0.01,
pstc_in=0.01,
pstc_gate=0.01,
invert_first=False,
invert_second=False,
mode="default",
output_scale=1,
):
if isinstance(A, str):
A = self.network.getNode(A)
if isinstance(B, str):
B = self.network.getNode(B)
if isinstance(C, str):
C = self.network.getNode(C)
dimensions = C.dimension
if (B is not None and B.dimension != dimensions) or (A is not None and A.dimension != dimensions):
raise Exception("Dimensions not the same for convolution (%d,%d->%d)" % (A.dimension, B.dimension, C.dimension))
if mode == "direct":
D = DirectConvolution(name, dimensions, invert_first, invert_second)
self.add(D)
D.getTermination("A").setTau(pstc_in)
D.getTermination("B").setTau(pstc_in)
D.getTermination("gate").setTau(pstc_gate)
if A is not None:
self.connect(A, D.getTermination("A"))
if B is not None:
self.connect(B, D.getTermination("B"))
self.connect(D.getOrigin("C"), C, pstc=pstc_out, weight=output_scale)
else:
D = make_array(self, name, N_per_D, dimensions, quick=quick, encoders=encoders, radius=radius)
A2 = input_transform(dimensions, True, invert_first)
B2 = input_transform(dimensions, False, invert_second)
D.addDecodedTermination("A", A2, pstc_in, False)
D.addDecodedTermination("B", B2, pstc_in, False)
if A is not None:
self.connect(A, D.getTermination("A"))
if B is not None:
self.connect(B, D.getTermination("B"))
ifftm2 = output_transform(dimensions)
self.connect(D, C, func=product, transform=ifftm2 * output_scale, pstc=pstc_out)
if self.network.getMetaData("convolution") == None:
self.network.setMetaData("convolution", HashMap())
bindings = self.network.getMetaData("convolution")
binding = HashMap(15)
binding.put("name", name)
if A is not None:
binding.put("A", A.getName())
else:
binding.put("A", None)
if B is not None:
binding.put("B", B.getName())
else:
binding.put("B", None)
binding.put("C", C.getName())
binding.put("N_per_D", N_per_D)
binding.put("quick", quick)
binding.put("encoders", MU.clone(encoders))
binding.put("radius", radius)
binding.put("pstc_out", pstc_out)
binding.put("pstc_in", pstc_in)
binding.put("pstc_gate", pstc_gate)
binding.put("invert_first", invert_first)
binding.put("invert_second", invert_second)
binding.put("mode", mode)
binding.put("output_scale", output_scale)
bindings.put(name, binding)
if self.network.getMetaData("templates") == None:
self.network.setMetaData("templates", ArrayList())
templates = self.network.getMetaData("templates")
templates.add(name)
if self.network.getMetaData("templateProjections") == None:
self.network.setMetaData("templateProjections", HashMap())
templateproj = self.network.getMetaData("templateProjections")
if A is not None:
templateproj.put(name, A.getName())
if B is not None:
templateproj.put(name, B.getName())
templateproj.put(name, C.getName())
return D
def getValue(self):
return MU.clone(self.matrix)
def make_convolution(self,
name,
A,
B,
C,
N_per_D,
quick=False,
encoders=[[1, 1], [1, -1], [-1, 1], [-1, -1]],
radius=3,
pstc_out=0.01,
pstc_in=0.01,
pstc_gate=0.01,
invert_first=False,
invert_second=False,
mode='default',
output_scale=1):
if isinstance(A, str):
A = self.network.getNode(A)
if isinstance(B, str):
B = self.network.getNode(B)
if isinstance(C, str):
C = self.network.getNode(C)
dimensions = C.dimension
if (B is not None
and B.dimension != dimensions) or (A is not None
and A.dimension != dimensions):
raise Exception('Dimensions not the same for convolution (%d,%d->%d)' %
(A.dimension, B.dimension, C.dimension))
if mode == 'direct':
D = DirectConvolution(name, dimensions, invert_first, invert_second)
self.add(D)
D.getTermination('A').setTau(pstc_in)
D.getTermination('B').setTau(pstc_in)
D.getTermination('gate').setTau(pstc_gate)
if A is not None:
self.connect(A, D.getTermination('A'))
if B is not None:
self.connect(B, D.getTermination('B'))
self.connect(D.getOrigin('C'), C, pstc=pstc_out, weight=output_scale)
else:
D = make_array(self,
name,
N_per_D,
dimensions,
quick=quick,
encoders=encoders,
radius=radius)
A2 = input_transform(dimensions, True, invert_first)
B2 = input_transform(dimensions, False, invert_second)
D.addDecodedTermination('A', A2, pstc_in, False)
D.addDecodedTermination('B', B2, pstc_in, False)
if A is not None:
self.connect(A, D.getTermination('A'))
if B is not None:
self.connect(B, D.getTermination('B'))
ifftm2 = output_transform(dimensions)
self.connect(D,
C,
func=product,
transform=ifftm2 * output_scale,
pstc=pstc_out)
if self.network.getMetaData("convolution") == None:
self.network.setMetaData("convolution", HashMap())
bindings = self.network.getMetaData("convolution")
binding = HashMap(15)
binding.put("name", name)
if A is not None:
binding.put("A", A.getName())
else:
binding.put("A", None)
if B is not None:
binding.put("B", B.getName())
else:
binding.put("B", None)
binding.put("C", C.getName())
binding.put("N_per_D", N_per_D)
binding.put("quick", quick)
binding.put("encoders", MU.clone(encoders))
binding.put("radius", radius)
binding.put("pstc_out", pstc_out)
binding.put("pstc_in", pstc_in)
binding.put("pstc_gate", pstc_gate)
binding.put("invert_first", invert_first)
binding.put("invert_second", invert_second)
binding.put("mode", mode)
binding.put("output_scale", output_scale)
bindings.put(name, binding)
if self.network.getMetaData("templates") == None:
self.network.setMetaData("templates", ArrayList())
templates = self.network.getMetaData("templates")
templates.add(name)
if self.network.getMetaData("templateProjections") == None:
self.network.setMetaData("templateProjections", HashMap())
templateproj = self.network.getMetaData("templateProjections")
if A is not None:
templateproj.put(name, A.getName())
if B is not None:
templateproj.put(name, B.getName())
templateproj.put(name, C.getName())
return D
