ReduceScaleValueTransform

class maze.train.trainers.common.value_transform.ReduceScaleValueTransform(epsilon: float)

Scale reduction value transform according to Pohlen et al (2018).

Parameters:

epsilon – Lipschitz constant for value function scaling.

transform_value(x: float | numpy.ndarray | torch.Tensor) → float | numpy.ndarray | torch.Tensor

(overrides ValueTransform)

implementation of ValueTransform interface

transform_value_inv(x: float | numpy.ndarray | torch.Tensor) → float | numpy.ndarray | torch.Tensor

(overrides ValueTransform)

implementation of ValueTransform interface