Convergence of a General Stochastic Approximation Process under Convex Constraints and some Applications

Albert and Gardner applied in [1] the stochastic approximation methods to the esti-mation of the vector parameter θ (θ ∈ ℝn) of a regression model $$ {v_{n}} = {g_{n}}\left( \theta \right) + {r_{n}} $$ where for n ≥ 1, vn is an observable random variable in ∝, gn a known function from ∝k into ∝, rn a random variable in ∝ whose expectation is 0.