RangeZeroOneObservationNormalizationStrategy¶
-
class
maze.core.wrappers.observation_normalization.normalization_strategies.range_zero_one.
RangeZeroOneObservationNormalizationStrategy
(observation_space: gym.spaces.Box, clip_range: Tuple[Union[float, int], Union[float, int]], axis: Optional[Union[int, Tuple[int], List[int]]])¶ Normalizes observations to value range [0, 1].
The strategy subtracts in a first step the minimum observed value to shift the lowest value after normalization to zero. In a subsequent step we divide the observation with the maximum of the previous step yielding observations in the range [0, 1].
-
estimate_stats
(observations: List[numpy.ndarray]) → Dict[str, Union[numpy.ndarray, float, int, Iterable[Union[float, int]]]]¶ (overrides
ObservationNormalizationStrategy
)- Implementation of
ObservationNormalizationStrategy
interface.
-
normalize_value
(value: numpy.ndarray) → numpy.ndarray¶ (overrides
ObservationNormalizationStrategy
)- Implementation of
ObservationNormalizationStrategy
interface.
-