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