Monte Carlo algorithm for pricing American/Bermudan options. At each exercise date, regress the discounted continuation value on basis functions of the state (1, S, S^2, ldots or Laguerre polynomials) across paths, then exercise when immediate payoff exceeds the fitted continuation value. Bridges the dimension wall of PDE methods (which scale exponentially with factors) — LSM scales linearly, making it the standard for high-dimensional Bermudans (callable structured notes, autocallables).