Volume Sampling and Spectral Equalization in Randomized Alternating Projections (Q7374651)

We study alternating projection methods with volume sampled subspaces, deriving explicit performance bounds through a spectral equalization mechanism that connects volume sampling to the Cayley-Hamilton theorem. We introduce a computationally efficient uniform sampling scheme achieving comparable guarantees with an unexpected relaxation strategy.