A Space Mapping approach for the calibration of financial models with the application to the Heston model
Anna Clevenhaus, Claudia Totzeck, Matthias Ehrhardt · 2025
Read A Space Mapping approach for the calibration of financial models with the application to the Heston model by Anna Clevenhaus, Claudia Totzeck, Matthias Ehrhardt (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-14521.txt.
Introduction
Anna Clevenhaus, Claudia Totzeck, and Matthias Ehrhardt wrote a 2025 paper on fitting finance models to market prices. Their test case is the Heston model, a standard model in which market volatility itself moves at random. Few of its settings can be read off prices directly, so fitting means solving a hard search problem with many local traps.
Their trick, called space mapping, pairs the true model with a cheaper stand-in. The stand-in prices a simpler option with a fast equation, while the true model prices an Asian option through a costlier route. Steps taken on the cheap model steer the fit of the true one. Test runs suggest the method can make such fits faster and more reliable, a daily need for trading desks.
Catalog record
| CREATOR | Anna Clevenhaus, Claudia Totzeck, Matthias Ehrhardt |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-14521
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Space Mapping approach for the calibration of financial models with the application to the Heston model Anna Clevenhaus, Claudia Totzeck, Matthias Ehrhardt Abstract We present a novel approach for parameter calibration of the Heston model for pricing an Asian put option, namely space mapping. Since few parameters of the Heston model can be directly extracted from real market data, calibration to real market data is implicit and therefore a challenging task. In addition, some of the parameters in the model are non-linear, which makes it difficult to find the global minimum of the optimization problem within the calibration. Our approach is based on the idea of space mapping, exploiting the residuum of a coarse surrogate model that allows optimization and a fine model that needs to be calibrated. In our case, the pricing of an Asian option using the Heston model SDE is the fine model, and the surrogate is chosen to be the Heston model PDE pricing a European option. We formally derive a gradient descent algorithm for the PDE constrained calibration model using well-known techniques from optimization with PDEs. Our main goal is to provide evidence that the space mapping approach can be useful in financial calibration tasks. Numerical results underline the feasibility of our approach. AMS classification: 65M06, 65K10, 91-08 Keywords: parameter calibration, Heston model, space mapping, surrogate model, Asian option 1 Introduction Financial models aim to approximate the real market behavior of various underlyings. Since most model parameters are implicit in the real market data, calibrating the model parameters to fit the real market data is challenging [ 7 , 13 , 19 , 21 , 22 , 25 ] . In this article, we consider the well-known Heston model [ 11 ] , a two-dimensional stochastic differential equation (SDE) that allows to simulate the behavior of stock prices. In particular, for a given variance of the option we calibrate four implicit parameters – the volatility of the variance, the mean inversion rate, the long-term mean and the correlation factor of the Brownian motions in the model equations for the asset and the variance – to the market data. Our goal is to provide a proof-of-concept, so we choose the simplest model with constant coefficients. However, we are aware that the Heston model can be extended in various ways, e.g. by considering time-dependent parameters [ 22 ] or by including additional processes, which leads to an increase in the difficulty of the calibration process [ 25 ] . Various calibration techniques for the Heston model are available in the literature. Indeed, in the setting of constant parameters and for very specific use cases, there are approaches based on the closed-form valuation formula of the Heston model [ 22 , 7 ] . These are fast and provide information about the global minimum. For more general cases, the stochastic nature of the Heston model allows for Monte Carlo optimization methods [ 25 ] , which can also be used to calibrate the stock price and variance. The Monte Carlo theory is well established, but the approaches are computationally expensive and do not provide information about global or local minima. Recently, calibration approaches using neural networks, deep learning strategies as well as parallel GPU implementation of the Heston model have been proposed [ 13 , 19 , 20 , 18 , 9 ] . The networks must be trained individually for each model, and training requires appropriate data. Again, there is no information about global or local minima. An advantage is that once the neural networks are trained, they can be evaluated quickly. Our contribution to this line of research is a calibration algorithm for the Heston model that is independent of a specific characteristic function and easily extendable to time-dependent parameters. The core of the algorithm is based on the space mapping [ 2 ] , an iterative procedure that minimizes the residuum of a fine and a coarse model. In fact, to calibrate the parameters of the fine model, the coarse model is optimized and the fine model is only evaluated. To demonstrate the proposed approach, we calibrate the Heston SDE model for pricing Asian put options (fine model) to real market data. The calibration uses only one given market price. In our setting, the coarse model is chosen because it is the computationally less expensive Heston model for pricing the European put option, which is known to be a good approximation for the Asian option in the case of short maturities. This is because the European option is a limiting case for the Asian option. In more detail, while we aim to calibrate the parameters of the SDE for the Asian put option, we will optimize the deterministic PDE model of the European option, which can be solved using techniques from optimization with PDE [ 12 , 27 ] . In each iteration of the calibration process, we will evaluate the SDE to compute the residuum of the two models. To our knowledge, there is only one paper dealing with the calibration of Asian options under the Heston model, where Khalife et al. [ 15 ] calibrate the Heston model using artificial intelligence and the approximation is done via the European option as well. The proposed calibration is based on aggressive space mapping (ASM) [ 8 ] . To iteratively update the parameters of the fine model, a PDE-constrained optimization problem is considered at the level of the coarse model. We measure the difference between the fair price given by the numerical solution of our model and the reference data, the subsequent market data, in the cost functional. The PDE-constrained optimization problem is solved using a gradient descent method. For this purpose, we formally derive an adjoint-based gradient descent algorithm for the Heston PDE model. Note that gradient descent algorithms have previously been used implicitly in the context of neural network approximations for the Heston model [ 20 ] . As mentioned above, our main goal is to provide a proof-of-concept that the space mapping approach can be useful for calibration tasks in financial research. This claim is supported by the numerical results. However, we are aware that there is much room for improvement, for example an extension to time-dependent parameters, see [ 5 ] for details on the PDE-constrained optimization problem. The article has the following structure. In Section 2 , we present the mathematical formulation of the Heston SDE model and its \log -transformed form for pricing an Asian put option. We also discuss the PDE formulation of the Heston model for pricing a European put option. In Section 3 , we then introduce the corresponding optimization problems and present details of the space mapping approach. This includes the formal derivation of the adjoint of the \log -transformed Heston model as well as the identification of the gradient which is used in the descent algorithm. We then discretize the SDE model and the PDE calibration method in Section 4 . Numerical results to illustrate the feasibility of space mapping in finance are presented in Section 5 . Finally, in Section 6 the paper ends with a conclusion and an outlook. 2 The Heston SDE model and its PDE approximation The Heston model