VIP

class pymc_experimental.model.transforms.autoreparam.VIP(_logit_lambda: Dict[str, TensorSharedVariable])

Helper to reparemetrize VIP model.

Manipulation of \(\lambda\) in the below equation is done using this helper class.

\[\begin{split}\begin{align*} \eta_{k} &\sim \text{normal}(\lambda_{k} \cdot \mu, \sigma^{\lambda_{k}})\\ \theta_{k} &= \mu + \sigma^{1 - \lambda_{k}} ( \eta_{k} - \lambda_{k} \cdot \mu) \sim \text{normal}(\mu, \sigma). \end{align*}\end{split}\]

__init__(_logit_lambda: Dict[str, TensorSharedVariable]) → None

Methods

__init__(_logit_lambda)
fit(*args, **kwargs)	Set \(\lambda_k\) using Variational Inference.
get_lambda()	Get \(\lambda_k\) that are currently used by the model.
set_all_lambda(value)	Set \(\lambda_k\) globally.
set_lambda(**kwargs)	Set \(\lambda_k\) per variable.
truncate_all_lambda(value)	Truncate all \(\lambda_k\) with \(\varepsilon\).
truncate_lambda(**kwargs)	Truncate \(\lambda_k\) with \(\varepsilon\).

Attributes

variational_parameters

Return raw \(\operatorname{logit}(\lambda_k)\) for custom optimization.