Transform#

class pymc_marketing.terms.Transform(inner, func)[source]#

Apply a pytensor function to a term’s output.

Delegates coordinates, data registration, and data updating to the inner expression. Covers link functions and arbitrary transformations without a separate link system.

Parameters:

innerAny: Inner expression accepted by build_param (ModelTerm, Sum, float, Prior, etc.).
funcCallable: A pytensor function applied to build_param(inner). E.g., pytensor.xtensor.math.exp, ptx.math.sigmoid.

Examples

import pytensor.xtensor as ptx

sigma = Transform(Intercept(name="sigma"), func=ptx.math.exp)

inner may be any expression, including custom ModelTerm subclasses and composed trees:

from pymc_extras.prior import Prior
from pymc_marketing.terms import Dot, Transform

effect = Transform(
    Dot(var_name="x", prior=Prior("Normal", dims="feature")),
    func=ptx.math.softplus,
)

Methods

`Transform.__init__`(inner, func)
`Transform.add_coords`(ds)	Add coordinates to the model during `register_data`.
`Transform.create_variable`()	Build `func(build_param(inner))`.
`Transform.from_dict`(data)	Reconstruct a transform from its serialized form.
`Transform.get_coords`(ds)	Collect coordinates from the inner expression.
`Transform.register_data`(ds)	Register shared data for the inner expression.
`Transform.set_data`(ds[, model])	Update shared data for the inner expression.
`Transform.to_dict`([_orig])

Attributes

`Transform.inner`
`Transform.func`