Great and thought provoking article! I immediately thought of “spread spectrum techniques”. In contrast with radio transmission techniques that use a single communication frequency, “spread spectrum” techniques spread the signal across a wide array of frequencies.
Somebody trying to intercept the communication may not even be able to distinguish the signal from background noise. The only way to detect it is to know how to invert the spreading function.
In general I take issue with terms like “sum-max”, the inverse of the spreading function might not be a sum, it could be arbitrary and nonlinear. I bet your neural network examples are nonlinear.
Great and thought provoking article! I immediately thought of “spread spectrum techniques”. In contrast with radio transmission techniques that use a single communication frequency, “spread spectrum” techniques spread the signal across a wide array of frequencies.
Somebody trying to intercept the communication may not even be able to distinguish the signal from background noise. The only way to detect it is to know how to invert the spreading function.
https://en.wikipedia.org/wiki/Spread_spectrum
In general I take issue with terms like “sum-max”, the inverse of the spreading function might not be a sum, it could be arbitrary and nonlinear. I bet your neural network examples are nonlinear.
That’s fair. Open to suggested alternatives. I suggested alternatives that don’t presume a summation, but none seem clearly great to me.