Reference / binference.priors.LogisticNormal

binference.priors.LogisticNormal

class binference.priors.LogisticNormal(mean=None, cov=None, scale=None)[source]

Logistic-normal prior on the population fractions.

\(\eta \sim \mathcal N(\mu, \Sigma)\) and \(\rho = \mathrm{softmax}(\eta)\).

Parameters:
meanarray_like (K,), optional

Prior mean of the log-fractions (up to a constant), e.g. the log of reference fractions. Defaults to a density flat in the measure coordinate, so the prior acts on the log-density whatever the binning.

covarray_like (K, K) or float, optional

Covariance of \(\eta\): a positive variance, or a symmetric matrix that is positive semi-definite on zero-sum vectors. Only that zero-sum part matters, since adding a constant to \(\eta\) leaves \(\rho\) unchanged. None means i.i.d. components with standard deviation scale.

scaleHyperprior or float

Multiplies the standard deviations. Defaults to HalfNormal(2.0) (inferred) when cov is None and to Fixed(1.0) otherwise.

__init__(mean=None, cov=None, scale=None)[source]

Methods

__init__([mean, cov, scale])

bind(bins)

Attributes

kind

"fractions" (a distribution on the simplex) or "rates"