mxnet.gluon.probability.distributions.utils

Distribution utilities

Functions

constraint_check()	Unified check_constraint interface for both scalar and tensor
digamma()	Unified digamma interface for both scalar and tensor
erf()	Unified erf interface for both scalar and tensor
erfinv()	Unified erfinv interface for both scalar and tensor
gammaln()	Unified gammaln interface for both scalar and tensor
logit2prob(logit[, binary])	Convert logit into probability form.
prob2logit(prob[, binary])	Convert probability to logit form.
sample_n_shape_converter(size)	Convert size to the proper format for performing sample_n.
sum_right_most(x, ndim)	Sum along the right most ndim dimensions of x,

Classes

cached_property

alias of _CachedProperty

mxnet.gluon.probability.distributions.utils.cached_property

alias of _CachedProperty

mxnet.gluon.probability.distributions.utils.constraint_check()

Unified check_constraint interface for both scalar and tensor

mxnet.gluon.probability.distributions.utils.digamma()

Unified digamma interface for both scalar and tensor

mxnet.gluon.probability.distributions.utils.erf()

Unified erf interface for both scalar and tensor

mxnet.gluon.probability.distributions.utils.erfinv()

Unified erfinv interface for both scalar and tensor

mxnet.gluon.probability.distributions.utils.gammaln()

Unified gammaln interface for both scalar and tensor

mxnet.gluon.probability.distributions.utils.logit2prob(logit, binary=True)

Convert logit into probability form. For the binary case, sigmoid() is applied on the logit tensor. Whereas for the multinomial case, softmax is applied along the last dimension of the logit tensor.

mxnet.gluon.probability.distributions.utils.prob2logit(prob, binary=True)

Convert probability to logit form. For the binary case, the logit stands for log(p / (1 - p)). Whereas for the multinomial case, the logit denotes log(p).

mxnet.gluon.probability.distributions.utils.sample_n_shape_converter(size)

Convert size to the proper format for performing sample_n.