View source on GitHub |
Combiner for the Discretization preprocessing layer.
tf.keras.layers.experimental.preprocessing.Discretization.DiscretizingCombiner(
epsilon, num_bins
)
This class encapsulates the computations for finding the quantile boundaries of a set of data in a stable and numerically correct way. Its associated accumulator is a namedtuple('summaries'), representing summarizations of the data used to generate boundaries.
Attributes | |
---|---|
epsilon
|
Error tolerance. |
num_bins
|
The desired number of buckets. |
Methods
compute
compute(
values, accumulator=None
)
Compute a step in this computation, returning a new accumulator.
deserialize
deserialize(
encoded_accumulator
)
Deserialize an accumulator received from 'serialize()'.
extract
extract(
accumulator
)
Convert an accumulator into a dict of output values.
merge
merge(
accumulators
)
Merge several accumulators to a single accumulator.
restore
restore(
output
)
Create an accumulator based on 'output'.
serialize
serialize(
accumulator
)
Serialize an accumulator for a remote call.