Global average pooling operation for temporal data.
>>> input_shape = (2, 3, 4) >>> x = tf.random.normal(input_shape) >>> y = tf.keras.layers.GlobalAveragePooling1D()(x) >>> print(y.shape) (2, 4)
channels_first. The ordering of the dimensions in the inputs.
channels_lastcorresponds to inputs with shape
(batch, steps, features)while
channels_firstcorresponds to inputs with shape
(batch, features, steps).
False(default), the rank of the tensor is reduced for spatial dimensions. If
True, the temporal dimension are retained with length 1. The behavior is the same as for
(batch_size, steps)indicating whether a given step should be masked (excluded from the average).
data_format='channels_last': 3D tensor with shape:
(batch_size, steps, features)
data_format='channels_first': 3D tensor with shape:
(batch_size, features, steps)
keepdims=False: 2D tensor with shape
data_format='channels_last': 3D tensor with shape
(batch_size, 1, features)
data_format='channels_first': 3D tensor with shape
(batch_size, features, 1)