Comment by nabla9

In deep learning this is called local response normalization (LRN).

ConvNets use LRN where most active neurons inhibit other neurons at the same location in neighboring feature maps. http://www.cs.toronto.edu/~fritz/absps/imagenet.pdf