Tensorflow: Understanding tf.contrib.layers.instance_norm graph

Question

I'm trying to understand tf.contrib.layers.instance_norm graph:

according to this graph:

x = gamma * (x + x_mean) / x_std - beta

but it should be

x = gamma * (x - x_mean) / x_std + beta

I'm missing something?

Answer 1

I get it, it's x = gamma * (x - x_mean) / x_std + beta as it should be because in Sub order is beta - 'Sub op input'.