Variance Reduction Applied to Machine Learning for Pricing\n Bermudan/American Options in High Dimension

Ludovic Goudenège, Andrea Molent, Antonino Zanette

arXiv (Cornell University) · 2019 · 12 citations · 25 references

DOIFull text

Open access

Concepts

Abstract

In this paper we propose an efficient method to compute the price of\nmulti-asset American options, based on Machine Learning, Monte Carlo\nsimulations and variance reduction technique. Specifically, the options we\nconsider are written on a basket of assets, each of them following a\nBlack-Scholes dynamics. In the wake of Ludkovski's approach (2018), we\nimplement here a backward dynamic programming algorithm which considers a\nfinite number of uniformly distributed exercise dates. On these dates, the\noption value is computed as the maximum between the exercise value and the\ncontinuation value, which is obtained by means of Gaussian process regression\ntechnique and Monte Carlo simulations. Such a method performs well for low\ndimension baskets but it is not accurate for very high dimension baskets. In\norder to improve the dimension range, we employ the European option price as a\ncontrol variate, which allows us to treat very large baskets and moreover to\nreduce the variance of price estimators. Numerical tests show that the proposed\nalgorithm is fast and reliable, and it can handle also American options on very\nlarge baskets of assets, overcoming the problem of the curse of dimensionality.\n

References

25