SPA approximates gated activations for faster FHE inference
On [-5,5], SPA-balanced is 3.25× faster than MOAI for erf-GELU at comparable error.
2 stories this week · 3 this month · 3 in 6 months · mostly security
3 stories, newest first · one card per story
On [-5,5], SPA-balanced is 3.25× faster than MOAI for erf-GELU at comparable error.
The construction evaluates XOR by adding ciphertexts and all eight symmetric binary gates with two blind rotations.
A new operator stabilizes deep reinforcement learning when training runs under fully homomorphic encryption constraints.
Today 1.7K readers · 2.1K views · 350 reads · all time 110K views