Batchnorm

class tf_encrypted.layers.batchnorm.Batchnorm(input_shape: List[int], mean: numpy.ndarray, variance: numpy.ndarray, scale: numpy.ndarray, offset: numpy.ndarray, variance_epsilon: float = 1e-08)

Batch Normalization Layer

Parameters:

• input_shape (List[int]) – input shape of the data flowing into the layer
• mean (np.ndarray) –

  …

• variance (np.ndarray) –

  …

• scale (np.ndarray) –

  …

• offset (np.ndarray) –

  …

• variance_epsilon (float) –

  …

backward() → None

backward is not implemented for batchnorm

Raises:NotImplementedError

forward(x: tf_encrypted.protocol.pond.PondPrivateTensor) → tf_encrypted.protocol.pond.PondPrivateTensor

Forward pass for inference

get_output_shape() → List[int]

Returns the layer’s output shape