def __init__(self, inmod, outmod, learnParams=None, c0=None, **kwargs):
# 2nd half for int connections
kwargs['outSliceFrom'] = outmod.dim
kwargs['outSliceTo'] = outmod.dim*2
Connection.__init__(self, inmod, outmod, **kwargs)
if isinstance(outmod, GFNNLayer):
outmod.conns.append(self)
if learnParams is None:
learnParams = {'learn': True, 'w': 0.05, 'l': 0, 'm1': -1, 'm2': -50, 'e': 4, 'k': 1 } # Critical learning rule
if 'type' not in learnParams:
learnParams['type'] = 'allfreq'
self.setArgs(learnParams=learnParams, c0=c0)
self.setLearnParams(learnParams)
if c0 is not None:
self.c0 = np.zeros((self.outdim, self.indim), np.complex64)
self.c0[:] = c0
self.c = np.zeros((self.outdim, self.indim), np.complex64)
if inmod == outmod:
# Don't learn self-connections
# This could be inverted guassian too
self.mask = 1-eye(self.outdim, self.indim)
# ParameterContainer.__init__(self, self.indim*self.outdim)
self.k *= self.mask
self.kSteps = np.zeros((4, self.outdim, self.indim), np.complex64)
self.randomize()
self.reset()
def __init__(self, weight, maxIndex, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
assert self.indim == self.outdim, \
"Indim (%i) does not equal outdim (%i)" % (
self.indim, self.outdim)
self.weight = weight
self.maxIndex = maxIndex
def __init__(self, inmod, outmod, name=None,
inSliceFrom=0, inSliceTo=None, outSliceFrom=0, outSliceTo=None):
if outSliceTo is None:
outSliceTo = outmod.indim
size = outSliceTo - outSliceFrom
Connection.__init__(self, inmod, outmod, name,
inSliceFrom, inSliceTo, outSliceFrom, outSliceTo)
ParameterContainer.__init__(self, size)
def __init__(self, inmod, outmod, name=None,
inSliceFrom=0, inSliceTo=None, outSliceFrom=0, outSliceTo=None):
if inSliceTo is None:
inSliceTo = inmod.outdim
size = inSliceTo - inSliceFrom
Connection.__init__(self, inmod, outmod, name,
inSliceFrom, inSliceTo, outSliceFrom, outSliceTo)
ParameterContainer.__init__(self, size)
def __init__(self, inmod, outmod, permutation, blocksize, *args, **kwargs):
Connection.__init__(self, inmod, outmod, *args, **kwargs)
if self.indim != self.outdim:
raise ValueError("Indim (%i) does not equal outdim (%i)" % (
self.indim, self.outdim))
if len(permutation) * blocksize != self.indim:
raise ValueError(
"Permutation has wrong size: should be %i but is %i." %(
(self.indim / blocksize), len(permutation)))
self.permutation = array(permutation)
self.invpermutation = permute(list(range(len(permutation))), permutation)
self.blocksize = blocksize
def __init__(self, inmod, outmod, permutation, blocksize, *args, **kwargs):
Connection.__init__(self, inmod, outmod, *args, **kwargs)
if self.indim != self.outdim:
raise ValueError("Indim (%i) does not equal outdim (%i)" %
(self.indim, self.outdim))
if len(permutation) * blocksize != self.indim:
raise ValueError(
"Permutation has wrong size: should be %i but is %i." %
((self.indim / blocksize), len(permutation)))
self.permutation = array(permutation)
self.invpermutation = permute(range(len(permutation)), permutation)
self.blocksize = blocksize
def __init__(self, inmod, outmod, name=None,
inSliceFrom=0, inSliceTo=None, outSliceFrom=0, outSliceTo=None):
if inSliceTo is None:
inSliceTo = inmod.outdim
size = inSliceTo - inSliceFrom
# FIXME: call `super()` with named kwargs so that cooperative inhertiance will work, otherwise...
# >>> isinstance(LinearConnection, Connection)
# False
# >>> isinstance(LinearConnection, ParameterContainer)
# False
Connection.__init__(self, inmod, outmod, name,
inSliceFrom, inSliceTo, outSliceFrom, outSliceTo)
ParameterContainer.__init__(self, size)
def __init__(self, dx, dy, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
self.dx = dx
self.dy = dy
ParameterContainer.__init__(self, 4 * self.outdim)
for i in xrange(0, self.outdim):
self.params[2 + i * 4] = dx / 6.0
self.params[3 + i * 4] = dx / 4.0
self.xx = numpy.repeat([numpy.arange(0, self.dx, 1)], self.dy, axis=0).T
self.yy = numpy.repeat([numpy.arange(0, self.dy, 1)], self.dx, axis=0)
assert self.indim == self.dx * self.dy, "Indim (%i) does not equal dx * dy (%i %i)" % (
self.indim,
self.dx,
self.dy,
)
def __init__(self, inmod, outmod, name=None,
inSliceFrom=0, inSliceTo=None, outSliceFrom=0, outSliceTo=None):
size = inSliceTo - inSliceFrom
Connection.__init__(self, inmod, outmod, name,
inSliceFrom, inSliceTo, outSliceFrom, outSliceTo)
ParameterContainer.__init__(self, size)
def __init__(self, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
assert self.indim == self.outdim, \
"Indim (%i) does not equal outdim (%i)" % (
self.indim, self.outdim)
def __init__(self, *args, r=0.1, **kwargs):
Connection.__init__(self, *args, **kwargs)
ParameterContainer.__init__(self, self.indim * self.outdim)
self.r = r
self.dropoutMask = np.random.binomial(1, (1 - r), (self.indim))
def __init__(self, inmod, outmod, gain=1.0, **kwargs):
Connection.__init__(self, inmod, outmod, **kwargs)
self.gain = gain
self.setArgs(gain = self.gain)
def __init__(self, mother, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
self._replaceParamsByMother(mother)
def __init__(self, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
def __init__(self, inmod, outmod, **kwargs):
# 1st inputs are for external connection
kwargs['outSliceTo'] = outmod.dim
Connection.__init__(self, inmod, outmod, **kwargs)
def __init__(self, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
def __init__(self, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
ParameterContainer.__init__(self, self.indim * self.outdim)
def __init__(self, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
n = self.outdim # self.indim*self.outdim # connect only 1 input neuron to random output
# self.params = np.random.random_integers(0, 1, n)
self.params = np.ones(n)
def __init__(self, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
assert self.indim == self.outdim, \
"Indim (%i) does not equal outdim (%i)" % (
self.indim, self.outdim)
def __init__(self, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
n = self.outdim # self.indim*self.outdim # connect only 1 input neuron to random output
# self.params = np.random.random_integers(0, 1, n)
self.params = np.ones(n)
def __init__(self, mother, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
self._replaceParamsByMother(mother)
def __init__(self, *args, **kwargs):
Connection.__init__(self, *args, **kwargs)
ParameterContainer.__init__(self, self.indim*self.outdim)
