def __init__(self, input_dim, whitening_output_dim, output_dim, eps=0.001, **kwargs):
self.input_dim = input_dim
self.whitening_output_dim = whitening_output_dim
self.output_dim = output_dim
self.eps = eps
self.kwargs = kwargs
self.whiteningnode = WhiteningNode(input_dim, whitening_output_dim, **kwargs)
self.mcanode = MCANode(whitening_output_dim, output_dim, self.eps, **kwargs)
self.singlepcanode = CCIPCANode(whitening_output_dim, output_dim=1)
self.deMeanInput = self.kwargs.get('deMean', True)
self.whiteningnode.deMeanInput = False
self.mcanode.deMeanInput = False
self.singlepcanode.deMeanInput = False
self.xavg = signalAvgNode(mode=self.kwargs.get('avgMode', 'Avg'), avgN=self.kwargs.get('avgN',1000))
self.xderiv = signalDerivNode()
self.zbvar = signalVarNode()
self.T = self.kwargs.get('T', 1)
self._curreps = [self.eps for _ in xrange(output_dim)]
self._initExp = True
self.n = 1
self.v = np.zeros([output_dim, input_dim])
self.wv = np.zeros([whitening_output_dim, input_dim])
self.err = 0.0
self.derr = 0.0
self._newerr = 0.0
self._errcnt = 0
self._validTrainingModes = ['Incremental']
def __init__(self, input_dim, output_dim, **kwargs):
self.input_dim = input_dim
self.output_dim = output_dim
self.kwargs = kwargs
self.var_rel = self.kwargs.get('var_rel', 0.001)
self.beta = self.kwargs.get('beta', 1.1)
self.xavg = signalAvgNode(mode=self.kwargs.get('avgMode', 'Avg'),
avgN=self.kwargs.get('avgN', 1000))
self.deMeanInput = self.kwargs.get('deMean', True)
self.reduce = self.kwargs.get('reduce', False)
self.n = 1 # n value for the ccipca
self._v = 0.1 * np.random.randn(
self.output_dim,
self.input_dim) # Internal Eigen Vector (unNormalized)
self._d = np.sum(np.absolute(self._v)**2,
axis=-1)**(1. / 2) # Internal Eigen Values
self._vn = self._v / self._d.reshape(
self._v.shape[0], 1) # Internal Eigen Vector (Normalized)
self.explained_var_tot = self._d.sum() # Total Explained Variance
self.v = self._vn.copy(
) # Eigen Vector (Normalized) (reduced if reduce is True)
self.d = self._d.copy() # Eigen Value (reduced if reduce is True)
self.reducedDim = self.output_dim
self._validTrainingModes = ['Incremental']
def __init__(self, input_dim, whitening_output_dim, output_dim, eps=0.001, **kwargs):
self.input_dim = input_dim
self.whitening_output_dim = whitening_output_dim
self.output_dim = output_dim
self.eps = eps
self.kwargs = kwargs
self.whiteningnode = WhiteningNode(input_dim, whitening_output_dim, **kwargs)
self.mcanode = MCANode(whitening_output_dim, output_dim, self.eps, **kwargs)
self.singlepcanode = CCIPCANode(whitening_output_dim, output_dim=1)
self.deMeanInput = self.kwargs.get('deMean', True)
self.whiteningnode.deMeanInput = False
self.mcanode.deMeanInput = False
self.singlepcanode.deMeanInput = False
self.xavg = signalAvgNode(mode=self.kwargs.get('avgMode', 'Avg'), avgN=self.kwargs.get('avgN',1000))
self.xderiv = signalDerivNode()
self.zbvar = signalVarNode()
self.T = self.kwargs.get('T', 1)
self._curreps = [self.eps for _ in range(output_dim)]
self._initExp = True
self.n = 1
self.v = np.zeros([output_dim, input_dim])
self.wv = np.zeros([whitening_output_dim, input_dim])
self.err = 0.0
self.derr = 0.0
self._newerr = 0.0
self._errcnt = 0
self._validTrainingModes = ['Incremental']
def __init__(self, input_dim, output_dim, eps=0.001, gamma=1.0, **kwargs):
self.input_dim = input_dim
self.output_dim = output_dim
self.eps = eps
self.gamma = gamma
self.kwargs = kwargs
self.deMeanInput = self.kwargs.get('deMean', True)
self.xavg = signalAvgNode(mode=self.kwargs.get('avgMode', 'Avg'),
avgN=self.kwargs.get('avgN', 1000))
self.normalize = self.kwargs.get('normalize', True)
self.n = 1 # n value for the mca
self._v = 0.1 * np.random.randn(
self.output_dim,
self.input_dim) # Internal Eigen Vector (unNormalized)
_d = np.sum(np.absolute(self._v)**2,
axis=-1)**(1. / 2) # Internal Eigen Values
self._v = self._v / _d.reshape(self._v.shape[0],
1) # Internal Eigen Vector (Normalized)
self.v = self._v.copy(
) # Eigen Vector (Normalized) (reduced if reduce is True)
self.d = _d.copy() # Eigen Value (reduced if reduce is True)
self._validTrainingModes = ['Incremental']
def __init__(self, input_dim, output_dim, eps=0.001, gamma=1.0, **kwargs):
self.input_dim = input_dim
self.output_dim = output_dim
self.eps = eps
self.gamma = gamma
self.kwargs = kwargs
self.deMeanInput = self.kwargs.get('deMean', True)
self.xavg = signalAvgNode(mode=self.kwargs.get('avgMode', 'Avg'), avgN=self.kwargs.get('avgN',1000))
self.normalize = self.kwargs.get('normalize', True)
self.n = 1 # n value for the mca
self._v = 0.1*np.random.randn(self.output_dim, self.input_dim) # Internal Eigen Vector (unNormalized)
_d = np.sum(np.absolute(self._v)**2,axis=-1)**(1./2) # Internal Eigen Values
self._v = self._v/_d.reshape(self._v.shape[0],1) # Internal Eigen Vector (Normalized)
self.v = self._v.copy() # Eigen Vector (Normalized) (reduced if reduce is True)
self.d = _d.copy() # Eigen Value (reduced if reduce is True)
self._validTrainingModes = ['Incremental']
def __init__(self, input_dim, output_dim, **kwargs):
self.input_dim = input_dim
self.output_dim = output_dim
self.kwargs = kwargs
self.var_rel = self.kwargs.get('var_rel',0.001)
self.beta = self.kwargs.get('beta',1.1)
self.xavg = signalAvgNode(mode=self.kwargs.get('avgMode', 'Avg'), avgN=self.kwargs.get('avgN',1000))
self.deMeanInput = self.kwargs.get('deMean', True)
self.reduce = self.kwargs.get('reduce', False)
self.n = 1 # n value for the ccipca
self._v = 0.1*np.random.randn(self.output_dim, self.input_dim) # Internal Eigen Vector (unNormalized)
self._d = np.sum(np.absolute(self._v)**2,axis=-1)**(1./2) # Internal Eigen Values
self._vn = self._v/self._d.reshape(self._v.shape[0],1) # Internal Eigen Vector (Normalized)
self.explained_var_tot = self._d.sum() # Total Explained Variance
self.v = self._vn.copy() # Eigen Vector (Normalized) (reduced if reduce is True)
self.d = self._d.copy() # Eigen Value (reduced if reduce is True)
self.reducedDim = self.output_dim
self._validTrainingModes = ['Incremental']