pymc.dims.WeightedZeroSumNormal

class pymc.dims.WeightedZeroSumNormal(*args, weights=None, core_dims=None, dims=None, default_transform=UNSET, observed=None, **kwargs)

Weighted zero-sum multivariate normal distribution.

Draws satisfy (weights * value).sum(core_dim) = 0. Generalization of ZeroSumNormal; with equal weights the two coincide. Exactly one core dimension is supported.

Parameters:

sigmaxtensor_like, optional

The standard deviation of the underlying unconstrained normal distribution. Defaults to 1.0. It cannot have core dimensions.

weightsxtensor_like

Strictly positive weights along the single core dimension.

core_dimsstr or Sequence of str

The single dimension along which the constraint is applied.

**kwargs

Additional keyword arguments used to define the distribution.

Returns:

XTensorVariable

An xtensor variable representing the weighted zero-sum normal distribution.

Methods

WeightedZeroSumNormal.dist([sigma, weights, ...])