Paper: arXiv 2311.07211
Authors: Jirong Zhuang, Deng Ding, Weiguo Lu, Xuan Wu, Gangnan Yuan
Abstract
In this work, we present a novel machine learning approach for pricing high-dimensional American options based on the modified Gaussian process regression (GPR). We incorporate deep kernel learning and sparse variational Gaussian processes to address the challenges traditionally associated with GPR. These challenges include its diminished reliability in high-dimensional scenarios and the excessive computational costs associated with processing extensive numbers of simulated paths Our findings indicate that the proposed method surpasses the performance of the least squares Monte Carlo method in high-dimensional scenarios, particularly when the underlying assets are modeled by Merton’s jump diffusion model. Moreover, our approach does not exhibit a significant increase in computational time as the number of dimensions grows. Consequently, this method emerges as a potential tool for alleviating the challenges posed by the curse of dimensionality.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 6.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper involves advanced mathematical concepts like Gaussian Processes, deep kernel learning, and variational inference, with detailed derivations in the theory sections. It also presents empirical experiments comparing performance against Least Squares Monte Carlo under jump diffusion models, discussing computational time and curse of dimensionality, indicating a backtest-ready approach.
Research Flowchart
flowchart TD A["Research Goal<br>Pricing High-dim American Options"] --> B["Data & Model Setup<br>Simulated Paths, Merton Jump Diffusion"] B --> C["Core Methodology<br>Modified Gaussian Process Regression"] C --> D["Key Innovations<br>Deep Kernel Learning & Sparse Variational GPs"] D --> E["Computational Process<br>Addressing Curse of Dimensionality"] E --> F["Key Findings<br>Outperforms LS Monte Carlo"] F --> G["Outcome<br>Scalable, Computationally Efficient"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a method for pricing high-dimensional American options that combines the Longstaff‑Schwartz algorithm with Deep Kernel Learning (DKL).
Source sentence
To address the above two challenges, we propose a method for pricing high-dimensional American options that combines the Longstaff-Schwartz algorithm with Deep Kernel Learning (DKL) proposed by Wilson et al. (2016a,b), which is a hybrid of Gaussian process and deep neural networks.
Assumes 5
- Underlying assets follow Merton's jump diffusion model. model
Source sentence
particularly when the underlying assets are modeled by Merton's jump diffusion model.
- Underlying assets follow geometric Brownian motion (GBM) in some experiments. model
Source sentence
... under geometric Brownian motion (GBM) and Merton’s jump diffusion model (MJD).
- Gaussian process prior with zero mean is used for the regression model. statistical
Source sentence
In this paper, we consider only the Gaussian process prior with zero mean, denoted as f ∼GP(0, k).
- Training data consist of simulated independent price paths. data
Source sentence
Simulate N independent paths
Show 1 more
- A sparse variational Gaussian process reduces computational cost. statistical
Source sentence
variational inference (Titsias, 2009; Hensman et al., 2013) is applied to reduce the computational cost.
Data
- Synthetic GBM and Merton’s jump diffusion paths synthetic unknown
Source sentence
we introduce a suite of numerical experiments designed to evaluate the efficacy of Deep Kernel Learning (DKL) for the pricing of American options under geometric Brownian motion (GBM) and Merton’s jump diffusion model (MJD).
Costs
Not applicable
Finds 4
- The proposed DKL method outperforms the least‑squares Monte Carlo method in high‑dimensional scenarios. performance
Source sentence
Our findings indicate that the proposed method surpasses the performance of the least squares Monte Carlo method in high-dimensional scenarios.
- DKL models show no significant increase in computation time as dimensions grow, indicating resistance to the curse of dimensionality. performance
Source sentence
It is noteworthy that there are no significant increases in computation time as the dimensions increase, leading to the conclusion that DKL models are not susceptible to the curse of dimensionality.
- With enough inducing points (M ≥40), DKL is both faster and more accurate than conventional GPR. performance
Source sentence
DKL models with a sufficient number of inducing points (M ≥40) are both faster and more accurate than GPR.
- DKL outperforms least‑squares regression without requiring specific hyper‑parameter selection across scenarios. performance
Source sentence
high-dimensional settings, and doesn’t required specific selection of hyper-parameters in different scenarios.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 22 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (29)
- Valuation of the early-exercise price for options using simulations and nonparametric regression (1996)
- Regression methods for pricing complex American-style options (2001)
- Valuing American options by simulation: a simple least-squares approach (2001)
- Kernel-based Monte Carlo simulation for American option pricing (2009)
- Pricing high-dimensional American options by kernel ridge regression (2020)
- American option pricing with machine learning: An extension of the Longstaff-Schwartz method (2021)
- Pricing of High-Dimensional American Options by Neural Networks (2010)
- Neural network regression for Bermudan option pricing (2021)
and 21 more
- Pricing Bermudan options using regression trees/random forests (2023)
- Machine learning for quantitative finance: fast derivative pricing, hedging and fitting (2018)
- Gaussian process regression for derivative portfolio modeling and application to credit valuation adjustment computations (2020)
- KrigHedge: Gaussian process surrogates for delta hedging (2021)
- Kriging metamodels and experimental design for Bermudan option pricing (2018)
- Variance reduction applied to machine learning for pricing Bermudan/American options in high dimension (2019) — link
- Machine learning for pricing American options in high-dimensional Markovian and non-Markovian models (2020)
- DGM: A deep learning algorithm for solving partial differential equations (2018)
- Deep kernel learning (2016)
- Stochastic variational deep kernel learning (2016)
- Variational learning of inducing variables in sparse Gaussian processes (2009)
- Gaussian Processes for Big Data (2013)
- Gaussian Processes for Machine Learning (2005) — link
- Universal Kernels. (2006)
- Gaussian processes and kernel methods: A review on connections and equivalences (2018) — link
- On the Surprising Behavior of Distance Metrics in High Dimensional Space (2001)
- Gpytorch: Blackbox matrix-matrix gaussian process inference with gpu acceleration (2018)
- Deep optimal stopping (2019)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library (2019)
- Scikit-learn: Machine learning in Python (2011)
- Scalable variational Gaussian process classification (2015)
0 of 29 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.