tfsnippet.preprocessing

class tfsnippet.preprocessing.BaseSampler

Bases: tfsnippet.dataflow.data_mappers.DataMapper

Base class for samplers.

_transform(x)

Subclasses should override this to implement the transformation.

sample(x)

Sample array according to x.

Parameters:x (np.ndarray) – The input x array. Returns:The sampled array. Return type:np.ndarray

class tfsnippet.preprocessing.BernoulliSampler(dtype=<type 'numpy.int32'>, random_state=None)

Bases: tfsnippet.preprocessing.samplers.BaseSampler

A DataMapper which can sample 0/1 integers according to the input probability. The input is assumed to be float numbers range within [0, 1) or [0, 1].

__init__(dtype=<type 'numpy.int32'>, random_state=None)

Construct a new BernoulliSampler.

Parameters:

• dtype – The data type of the sampled array. Default np.int32.
• random_state (RandomState) – Optional numpy RandomState for sampling. (default None, use the global RandomState).

dtype

Get the data type of the sampled array.

sample(x)

Sample array according to x.

Parameters:x (np.ndarray) – The input x array. Returns:The sampled array. Return type:np.ndarray

class tfsnippet.preprocessing.UniformNoiseSampler(minval=0.0, maxval=1.0, dtype=None, random_state=None)

Bases: tfsnippet.preprocessing.samplers.BaseSampler

A DataMapper which can add uniform noise onto the input array. The data type of the returned array will be the same as the input array, unless dtype is specified at construction.

__init__(minval=0.0, maxval=1.0, dtype=None, random_state=None)

Construct a new UniformNoiseSampler.

Parameters:

• minval – The lower bound of the uniform noise (included).
• maxval – The upper bound of the uniform noise (excluded).
• dtype – The data type of the sampled array. Default np.int32.
• random_state (RandomState) – Optional numpy RandomState for sampling. (default None, use the global RandomState).

dtype

Get the data type of the sampled array.

maxval

Get the upper bound of the uniform noise (excluded).

minval

Get the lower bound of the uniform noise (included).

sample(x)

Sample array according to x.

Parameters:x (np.ndarray) – The input x array. Returns:The sampled array. Return type:np.ndarray