cutmix does: replaces a patch of one image with a patch from another

gradient clipping does: caps gradient norm to prevent exploding gradients

Gradient clipping caps gradient norm to prevent exploding gradients

structured pruning removes: entire filters or attention heads, not individual weights

Structured pruning removes entire filters or attention heads, not individual weights

operator fusion does at the compiler level: merges adjacent ops to reduce memory traffic

Operator fusion merges adjacent operations to optimize execution and reduce memory traffic

aliasing is: high frequencies masquerading as low frequencies due to undersampling

Aliasing occurs when high frequencies masquerade as low frequencies due to undersampling

a low-pass filter does: removes frequencies above a cutoff, keeps slow-varying signal

Low-pass filter: removes frequencies above cutoff, retains slow-varying signal

label smoothing does: replaces one-hot [0,0,1,0] with [0.025, 0.025, 0.925, 0.025]

Label smoothing replaces hard labels with soft labels to regularize neural networks