Denoising Monte Carlo sensitivity estimates

Monte Carlo methods are in widespread use both in academia and industry. We are, in particular, interested in improving sensitivity estimates obtained from Monte Carlo experiments with respect to given parameter values, motivated by, but not restricted to, financial applications. Denoising and interpolation methods, which have been used for a long time in many different areas, are proposed in a new form which is quadratic, easy to implement, and tailored to our objectives. This heuristic approach is supported by numerical experiments.