jax.nn.initializers module
Contents
jax.nn.initializers module#
Common neural network layer initializers, consistent with definitionsused in Keras and Sonnet.
Initializers#
This module provides common neural network layer initializers,consistent with definitions used in Keras and Sonnet.
An initializer is a function that takes three arguments:(key,shape,dtype) and returns an array with dimensionsshape anddata typedtype. Argumentkey is a PRNG key (e.g. fromjax.random.key()), used to generate random numbers to initialize the array.
| Builds an initializer that returns arrays full of a constant |
| Builds an initializer for delta orthogonal kernels. |
| Builds a Glorot normal initializer (aka Xavier normal initializer). |
| Builds a Glorot uniform initializer (aka Xavier uniform initializer). |
| Builds a He normal initializer (aka Kaiming normal initializer). |
| Builds a He uniform initializer (aka Kaiming uniform initializer). |
| Builds a He normal initializer (aka Kaiming normal initializer). |
| Builds a He uniform initializer (aka Kaiming uniform initializer). |
| Builds a Lecun normal initializer. |
| Builds a Lecun uniform initializer. |
| Builds an initializer that returns real normally-distributed random arrays. |
| An initializer that returns a constant array full of ones. |
| Builds an initializer that returns uniformly distributed orthogonal matrices. |
| Builds an initializer that returns truncated-normal random arrays. |
| Builds an initializer that returns real uniformly-distributed random arrays. |
| Initializer that adapts its scale to the shape of the weights tensor. |
| Builds a Glorot normal initializer (aka Xavier normal initializer). |
| Builds a Glorot uniform initializer (aka Xavier uniform initializer). |
| An initializer that returns a constant array full of zeros. |
| Protocol for initializers returned by |
