• namespace: Rindow\NeuralNetworks\Layer
• classname: BatchNormalization

Batch Normalization layer.

Normalize the previous activation function layer at each batch

Methods

constructor

$builer->BatchNormalization(
    int $axis=-1,
    float $momentum=0.99,,
    float $epsilon=0.001,
    bool $center=true,
    bool $scale=true,
    string|callable $beta_initializer='zeros',
    string|callable $gamma_initializer='ones',
    string|callable $moving_mean_initializer='zeros',
    string|callable $moving_variance_initializer='ones',
    string $name=null,
)

You can create a BatchNormalization layer instances with the Layer Builder.

Options

• momentum: Momentum for the moving average.
  • default is 0.99
• epsilon: Small float added to variance to avoid dividing by zero.
  • default is 0.001
• beta_initializer: name of initializer for the beta weight.
  • default is zeros
• gamma_initializer: name of initializer for the gamma weight.
  • default is ones
• moving_mean_initializer: name of initializer for the moving mean.
  • default is zeros
• moving_variance_initializer: name of initializer for the moving variance.
  • default is ones

Examples

$model->add($nn->layers()->BatchNormalization(
    momentum:0.99,
    epsilon:0.001,
    beta_initializer:'zeros',
    gamma_initializer:'ones',
    moving_mean_initializer:'zeros',
    moving_variance_initializer:'ones',
));