transforms.Transform
Base class for named transforms with estimable parameters.
Usage
transforms.Transform()Subclasses must implement apply_pymc() and define name and param_specs. Stateful transforms (e.g. adstock) should also override step() and has_state.
Attributes
| Name | Description |
|---|---|
| has_state | Whether this transform carries state across time steps. |
has_state
Whether this transform carries state across time steps.
has_state: bool
Methods
| Name | Description |
|---|---|
| apply_pymc() | Apply the transform in the PyMC computation graph. |
| emit_prior() | Create a PyMC random variable for a transform parameter. |
| step() |
Apply one time step inside a pytensor.scan body.
|
apply_pymc()
Apply the transform in the PyMC computation graph.
Usage
apply_pymc(x, params, *, panel_info=None, data=None)emit_prior()
Create a PyMC random variable for a transform parameter.
Usage
emit_prior(param_name, spec)step()
Apply one time step inside a pytensor.scan body.
Usage
step(x_t, state, params)