Can you explain why "PV model reaches saturation faster than PI"?
why "the permutation-variant RISnet quickly converges while the permutation-invariant RISnet takes significantly longer time to train. . However, the achieved WSR is higher with the permutation-invariant RISnet"?
They are two different architectures. I designed them for different behaviors with permutation. It was unexpected that the convergence time and final performance are different.
Can you explain why "PV model reaches saturation faster than PI"? why "the permutation-variant RISnet quickly converges while the permutation-invariant RISnet takes significantly longer time to train. . However, the achieved WSR is higher with the permutation-invariant RISnet"?