tfp.experimental.marginalize.logeinsumexp
Stay organized with collections
Save and categorize content based on your preferences.
Computing einsum in logarithmic space.
tfp.experimental.marginalize.logeinsumexp(
formula, *tensors
)
Args |
|---|
formula
|
a formula compatible with einsum. (But note that
ellipses are not supported.)
|
*tensors
|
tensors to which contractions will be applied.
|
Returns |
|---|
logeinsumexp(formula, *tensors) is equivalent to
tf.math.log(einsum(formula, *map(tf.math.exp, tensors))) except
that it is more numerically stable.
|
Notes |
|---|
This implementation of logeinsumexp is primarily intended for internal
use in the marginalize library and assumes the formula is well-formed.
|
Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.
Last updated 2023-11-21 UTC.
[null,null,["Last updated 2023-11-21 UTC."],[],[],null,["# tfp.experimental.marginalize.logeinsumexp\n\n\u003cbr /\u003e\n\n|----------------------------------------------------------------------------------------------------------------------------------------------------------------|\n| [View source on GitHub](https://github.com/tensorflow/probability/blob/v0.23.0/tensorflow_probability/python/experimental/marginalize/logeinsumexp.py#L45-L64) |\n\nComputing `einsum` in logarithmic space. \n\n tfp.experimental.marginalize.logeinsumexp(\n formula, *tensors\n )\n\n\u003cbr /\u003e\n\n\u003cbr /\u003e\n\n\u003cbr /\u003e\n\n| Args ---- ||\n|------------|---------------------------------------------------------------------------------|\n| `formula` | a formula compatible with `einsum`. (But note that ellipses are not supported.) |\n| `*tensors` | tensors to which contractions will be applied. |\n\n\u003cbr /\u003e\n\n\u003cbr /\u003e\n\n\u003cbr /\u003e\n\n\u003cbr /\u003e\n\n| Returns ------- ||\n|---|---|\n| `logeinsumexp(formula, *tensors)` is equivalent to `tf.math.log(einsum(formula, *map(tf.math.exp, tensors)))` except that it is more numerically stable. ||\n\n\u003cbr /\u003e\n\n\u003cbr /\u003e\n\n\u003cbr /\u003e\n\n\u003cbr /\u003e\n\n| Notes ----- ||\n|---|---|\n| This implementation of `logeinsumexp` is primarily intended for internal use in the `marginalize` library and assumes the formula is well-formed. ||\n\n\u003cbr /\u003e"]]
View source on GitHub