Likelihoods

class tramp.likelihoods.GaussianLikelihood(y, var=1, y_name='y', isotropic=True)

class tramp.likelihoods.AbsLikelihood(y, y_name='y', isotropic=True)

class tramp.likelihoods.SgnLikelihood(y, y_name='y', isotropic=True)

class tramp.likelihoods.SymmetricDoorLikelihood(width, y, y_name='y', isotropic=True)

class tramp.likelihoods.ReluLikelihood(y, y_name='y', isotropic=True)

class tramp.likelihoods.LeakyReluLikelihood(slope, y, y_name='y', isotropic=True)

class tramp.likelihoods.HardTanhLikelihood(y, y_name='y', isotropic=True)

class tramp.likelihoods.HardSigmoidLikelihood(y, y_name='y', isotropic=True)

class tramp.likelihoods.AsymmetricAbsLikelihood(y, y_name='y', isotropic=True, shift=0.0001)

class tramp.likelihoods.ModulusLikelihood(y, y_name='y', isotropic=True)

Modulus likelihood \(y = |z|\).

Parameters

• y (ndarray) – observed modulus

• y_name (str) – name of y for display

Notes

For message passing it is more convenient to represent the complex array z as a real array Z where Z[0] = z.real and Z[1] = z.imag