m=1 n=3: 19999 converged sample points in [-4.0,4.0]^3 (from 20000 random starts); distinct sign vectors of x = 2; kNN(k=24)-graph clusters with edges < eps: eps=0.3: 2, eps=0.4: 2, eps=0.45: 2, eps=0.49: 2; predicted 2^m = 2; max |x0 - m| = 3.69e-07
m=2 n=5: 199995 converged sample points in [-4.0,4.0]^5 (from 200000 random starts); distinct sign vectors of x = 4; kNN(k=24)-graph clusters with edges < eps: eps=0.3: 4, eps=0.4: 4, eps=0.45: 4, eps=0.49: 4; predicted 2^m = 4; max |x0 - m| = 6.18e-07
m=3 n=7: 285154 converged sample points in [-4.0,4.0]^7 (from 400000 random starts); distinct sign vectors of x = 8; kNN(k=24)-graph clusters with edges < eps: eps=0.3: 271, eps=0.4: 20, eps=0.45: 11, eps=0.49: 9; predicted 2^m = 8; max |x0 - m| = 1.61e-06
Note: distinct predicted components are >= 2/R = 0.5 apart inside the box, so any eps < 0.5 cannot merge two of them; a cluster count above 2^m only reflects sparse sampling.
