Prior distribution#
We support Gaussian prior distributions,
specifically IsotropicGaussianPrior
and DiagonalGaussianPrior.
Alternatively, you can bring your own prior distribution by overwriting the abstract class
Prior. Please find the references below.
- class fortuna.prob_model.prior.base.Prior[source]#
Abstract prior distribution class.
- abstract log_joint_prob(params)[source]#
Evaluate the prior log-probability density function (a.k.a. log-pdf).
- Parameters:
params (PyTree) – The parameters where to evaluate the log-pdf.
- Returns:
Evaluation of the prior log-pdf.
- Return type:
float
- property rng: RandomNumberGenerator#
Invoke the random number generator object.
- Return type:
The random number generator object.
- abstract sample(params_like, rng=None)[source]#
Sample parameters from the prior distribution.
- Parameters:
params_like (PyTree) – An PyTree object with the same structure as the parameters to sample.
rng (Optional[PRNGKeyArray]) – A random number generator. If not passed, this will be taken from the attributes of this class.
- Returns:
A sample from the prior distribution.
- Return type:
PyTree
- class fortuna.prob_model.prior.gaussian.DiagonalGaussianPrior(log_var)[source]#
A diagonal Gaussian prior class.
- Parameters:
log_var (jnp.ndarray) – Prior log-variance vector corresponding to the logarithm of the diagonal of the prior covariance matrix.
- log_joint_prob(params)[source]#
Evaluate the prior log-probability density function (a.k.a. log-pdf).
- Parameters:
params (PyTree) – The parameters where to evaluate the log-pdf.
- Returns:
Evaluation of the prior log-pdf.
- Return type:
float
- property rng: RandomNumberGenerator#
Invoke the random number generator object.
- Return type:
The random number generator object.
- sample(params_like, rng=None)[source]#
Sample parameters from the prior distribution.
- Parameters:
params_like (PyTree) – An PyTree object with the same structure as the parameters to sample.
rng (Optional[PRNGKeyArray]) – A random number generator. If not passed, this will be taken from the attributes of this class.
- Returns:
A sample from the prior distribution.
- Return type:
PyTree
- class fortuna.prob_model.prior.gaussian.IsotropicGaussianPrior(log_var=0.0)[source]#
A diagonal Gaussian prior class.
- Parameters:
log_var (Optional[float]) – Prior log-variance value. The covariance matrix of the prior distribution is given by a diagonal matrix with this parameter on every entry of the diagonal.
- log_joint_prob(params)[source]#
Evaluate the prior log-probability density function (a.k.a. log-pdf).
- Parameters:
params (PyTree) – The parameters where to evaluate the log-pdf.
- Returns:
Evaluation of the prior log-pdf.
- Return type:
float
- property rng: RandomNumberGenerator#
Invoke the random number generator object.
- Return type:
The random number generator object.
- sample(params_like, rng=None)[source]#
Sample parameters from the prior distribution.
- Parameters:
params_like (PyTree) – An PyTree object with the same structure as the parameters to sample.
rng (Optional[PRNGKeyArray]) – A random number generator. If not passed, this will be taken from the attributes of this class.
- Returns:
A sample from the prior distribution.
- Return type:
PyTree