Callbacks: keras_proba.callbacks¶
Training callbacks.
ResetCovarianceCallback ¶
Bases: keras.src.callbacks.callback.Callback
Resets the precision matrix of every GP covariance layer at the start of each epoch.
RandomFeatureGaussianProcess accumulates its posterior precision matrix
over training minibatches. Since the feature extractor changes during
training, the estimate should be recomputed in each epoch, so that the
final covariance only reflects the last epoch's features [1].
Example:
[1]: Jeremiah Liu et al. Simple and Principled Uncertainty Estimation with Deterministic Deep Learning via Distance Awareness. In Neural Information Processing Systems, 2020. https://arxiv.org/abs/2006.10108
__doc__
class-attribute
¶
__doc__ = "Resets the precision matrix of every GP covariance layer at the start of each epoch.\n\n `RandomFeatureGaussianProcess` accumulates its posterior precision matrix\n over training minibatches. Since the feature extractor changes during\n training, the estimate should be recomputed in each epoch, so that the\n final covariance only reflects the last epoch's features [1].\n\n Example:\n\n ```python\n model.fit(x, y, epochs=10, callbacks=[ResetCovarianceCallback()])\n ```\n\n [1]: Jeremiah Liu et al. Simple and Principled Uncertainty Estimation with\n Deterministic Deep Learning via Distance Awareness.\n In _Neural Information Processing Systems_, 2020.\n https://arxiv.org/abs/2006.10108\n "
str(object='') -> str str(bytes_or_buffer[, encoding[, errors]]) -> str
Create a new string object from the given object. If encoding or errors is specified, then the object must expose a data buffer that will be decoded using the given encoding and error handler. Otherwise, returns the result of object.str() (if defined) or repr(object). encoding defaults to sys.getdefaultencoding(). errors defaults to 'strict'.
__module__
class-attribute
¶
str(object='') -> str str(bytes_or_buffer[, encoding[, errors]]) -> str
Create a new string object from the given object. If encoding or errors is specified, then the object must expose a data buffer that will be decoded using the given encoding and error handler. Otherwise, returns the result of object.str() (if defined) or repr(object). encoding defaults to sys.getdefaultencoding(). errors defaults to 'strict'.