rbnet.multivariate_normal.PairwiseProduct

class rbnet.multivariate_normal.PairwiseProduct(mean1, mean2, cov1=None, cov2=None, prec1=None, prec2=None)[source]

Bases: object

Public Data Attributes:

`cov1`
`cov2`
`prec1`
`prec2`
`sum_cov`
`sum_cov_inv`
`prec`
`cov`	Compute covariance matrix using the least expensive method

Public Methods:

`__init__`(mean1, mean2[, cov1, cov2, prec1, ...])
`mean`	Compute mean using the least expensive method
`torch`	Return the product distribution as torch MultivariateNormal
`log_norm`	Compute log-normalisation using the least expensive method

Private Methods:

`_cost_comp`(cost1, cost2, comp)	Compares cost1 and cost2 using comp. Costs are numpy arrays [i, mm, mv, ma, va] with i indicating the number of matrix inversion, mm that of matrix-matrix multiplications, mv that of matrix-vector multiplications, ma that of matrix-matrix additions, and va that of vector-vector additions. Comparison is performed element-wise in this order, that is::.
`_cost_lt`(cost1, cost2)
`_cov1_cost`()
`_cov2_cost`()
`_prec1_cost`()
`_prec2_cost`()
`_sum_cov_cost`()
`_sum_cov_inv_cost`()
`_prec_v1_cost`()	Cost of computing precision by adding prec1 and prec2
`_prec_v2_cost`()	Cost of computing precision by inverting cov (if cov is not present assume infinite costsfor tie-breaking).
`_prec_cost`()
`_cov_v1`()	Compute covariance from prec1 and prec2
`_cov_v1_cost`()
`_cov_v2`()	Compute covariance from cov1 and cov2
`_cov_v2_cost`()
`_cov_cost`()
`_mean_v1`()	Compute mean from cov, prec1, and prec2
`_mean_v1_cost`()
`_mean_v2`()	Compute mean from cov, prec1, and prec2
`_mean_v2_cost`()

property log_norm: Compute log-normalisation using the least expensive method

property torch: Return the product distribution as torch MultivariateNormal