chainer.functions.group_normalization(x, groups, gamma, beta, eps=1e-05)[source]

Group normalization function.

This function implements a “group normalization” which divides the channels into groups and computes within each group the mean and variance, then normalize by these statistics, scales and shifts them.

  • x (Variable or numpy.ndarray or cupy.ndarray) – Batch tensors. First dimension of this value must be the size of minibatch and second dimension must be the number of channels. Moreover, this value must have one or more following dimensions, such as height and width.
  • groups (int) – The number of channel groups. This value must be a divisor of the number of channels.
  • gamma (Variable) – Scaling parameter.
  • beta (Variable) – Shifting parameter.
  • eps (float) – Epsilon value for numerical stability of normalization.

The output variable which has the same shape as \(x\).

Return type:


See: Group Normalization