Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.8/site-packages/tensorflow_addons/layers/poincare.py: 46%
24 statements
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-03 07:57 +0000
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-03 07:57 +0000
1# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
2#
3# Licensed under the Apache License, Version 2.0 (the "License");
4# you may not use this file except in compliance with the License.
5# You may obtain a copy of the License at
6#
7# http://www.apache.org/licenses/LICENSE-2.0
8#
9# Unless required by applicable law or agreed to in writing, software
10# distributed under the License is distributed on an "AS IS" BASIS,
11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12# See the License for the specific language governing permissions and
13# limitations under the License.
14# ==============================================================================
15"""Implementing PoincareNormalize layer."""
17import tensorflow as tf
18from typeguard import typechecked
19from typing import Union, List
22@tf.keras.utils.register_keras_serializable(package="Addons")
23class PoincareNormalize(tf.keras.layers.Layer):
24 """Project into the Poincare ball with `norm <= 1.0 - epsilon`.
26 See [Poincaré Embeddings for Learning Hierarchical Representations](https://arxiv.org/pdf/1705.08039.pdf),
27 and [wiki](https://en.wikipedia.org/wiki/Poincare_ball_model).
29 For a 1-D tensor with `axis = 0`, computes
31 (x * (1 - epsilon)) / ||x|| if ||x|| > 1 - epsilon
32 output =
33 x otherwise
35 For `x` with more dimensions, independently normalizes each 1-D slice along
36 dimension `axis`.
38 Args:
39 axis: Axis along which to normalize. A scalar or a vector of integers.
40 epsilon: A small deviation from the edge of the unit sphere for
41 numerical stability.
42 """
44 @typechecked
45 def __init__(
46 self, axis: Union[None, int, List[int]] = 1, epsilon: float = 1e-5, **kwargs
47 ):
48 super().__init__(**kwargs)
49 self.axis = axis
50 self.epsilon = epsilon
52 def call(self, inputs):
53 x = tf.convert_to_tensor(inputs)
54 square_sum = tf.math.reduce_sum(tf.math.square(x), self.axis, keepdims=True)
55 x_inv_norm = tf.math.rsqrt(square_sum)
56 x_inv_norm = tf.math.minimum((1.0 - self.epsilon) * x_inv_norm, 1.0)
57 outputs = tf.math.multiply(x, x_inv_norm)
58 return outputs
60 def compute_output_shape(self, input_shape):
61 return input_shape
63 def get_config(self):
64 config = {"axis": self.axis, "epsilon": self.epsilon}
65 base_config = super().get_config()
66 return {**base_config, **config}