Autoencoder Option Pricing Models

Abstract

We propose a new framework allowing to estimate non-parametric affine and non-affine option pricing models. Our method applies autoencoder neural networks to the log-characteristic function implied by observed option prices. Since the log-characteristic function is linear in the state factors under the affine assumption, we obtain a data-driven affine model by specifying a linear mapping in the autoencoder architecture. Alternatively, we let the data speak about any needed non-linearities to estimate a non-affine model. Using an extensive panel of S&P 500 options, our approach reveals that the non-affine class of models only outperforms the affine class in pricing options out-of-sample when the number of factors is small and the prediction horizon is not far ahead into the future.

Keywords: Characteristic function, Affine jump-diffusion models, Latent factor models, Autoencoders, Option pricing

JEL Classification: C14, C58, G13