He Uniform Initialiser¶
he_uniform_init_type
he_uniform_init_type()
Draws samples from a uniform distribution designed for layers with ReLU activation.
\[W \sim U\left[-\sqrt{\frac{6}{n_{in}}}, \sqrt{\frac{6}{n_{in}}}\right]\]
where \(n_{in}\) is the number of input units.
This initialisation is specifically designed for networks using ReLU activations and helps prevent vanishing/exploding gradients.
Shape:¶
Initialises weights based on the shape provided during layer setup.