[source]

He

The He initializer was designed to initialize parameters that feed into rectified Activation layers such as those employing ReLU, Leaky ReLU, or ELU. It draws values from a uniform distribution with limits defined as +/- sqrt(6 / fanIn).

Note: In this library, Xavier 2 is a backward-compatible alias of He. They share the same scale and are interchangeable.

Parameters

This initializer does not have any parameters.

Example

use Rubix\ML\NeuralNet\Initializers\He;

$initializer = new He();

References

---

1. K. He et al. (2015). Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification. ↩