Paper presents runtime-throttleable neural networks for resource-constrained devices.
problem Resource constraints in edge computing platforms.
method Runtime-throttleable neural networks using block-level gating.
result Smooth performance throttling with minimal accuracy loss.