Deep Learning Option Pricing with Market Implied Volatility Surfaces
A study presents a learning pipeline to price options from SPX implied volatility surfaces. A variational autoencoder compresses each 820 point surface into a 10 dimensional code, and a multilayer perceptron maps code, strike, and maturity to prices for American puts and Asian options. Market surfaces are low dimensional, about 5 to 10 factors. Trained on 1,051 arbitrage free surfaces, it is GPU friendly, with errors at long maturities and mostly near at the money.
What it examines
The paper proposes a deep learning pricer that uses entire market-implied volatility surfaces. A VAE compresses SPX surfaces (2018--2023) into 10 latent factors; an MLP maps these with strike and maturity to prices for American puts and arithmetic Asian options. Training is staged then fine-tuned end-to-end using QuantLib labels.
What it concludes
The model prices American and Asian options accurately, with larger errors near long maturities and at-the-money. It offers fast, GPU-friendly, single-pass valuation for real-time risk and portfolio pricing, and can mimic proprietary engines. Limits: SPX focus, EOD data, grid choices. Future work: more assets, products, asset classes, and VAE-based surface generation.
Evidence objects
Researchers unveil an end-to-end learning pricer using full volatility surfaces to value exotics: a VAE compresses 820-point SPX surfaces to 10 dimensions; an MLP outputs put and arithmetic Asian prices.
key_findings bullet 1 · key_findings · validation V0
Trained on 1,051 arbitrage-free SPX surfaces from 2018--2023 with QuantLib labels, the system achieves high accuracy, with errors concentrated at long maturities and near at-the-money, where spreads dominate small prices.
key_findings bullet 2 · key_findings · validation V0
A notable insight: market surfaces are effectively low-dimensionalPCA and VAE suggest 5--10 factorsenabling single-pass, GPU-friendly valuation. Strengths: speed, scalability, adaptability. Limits: end-of-day SPX, constrained grids, engine-dependent labels, no-arbitrage, hedging/baselines comparisons.
key_findings bullet 3 · key_findings · validation V0
Links market-implied volatility surfaces to exotic option pricing using a VAE for compression and an MLP surrogate, trained on SPX EOD data with arbitrage-free filtering. Targets American and arithmetic Asian options, enabling faster pricing. Practical and relevant, but largely incrementalcommon architecture, QuantLib labels, staged training, limited novelty or guarantees, evidence.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
Abstract: We present a deep learning framework for pricing options based on market-implied volatility surfaces. Using end-of-day S\&P 500 index options quotes from 2018-2023, we construct arbitrage-free volatility surfaces and generate training data for American puts and arithmetic Asian options using QuantLib. To address the high dimensionality of volatility surfaces, we employ a variational autoencoder (V… ▽ More We present a deep learning framework for pricing options based on market-implied volatility surfaces. Using end-of-day S\&P 500 index options quotes from 2018-2023, we construct arbitrage-free volatility surfaces and generate training data for American puts and arithmetic Asian options using QuantLib. To address the high dimensionality of volatility surfaces, we employ a variational autoencoder (VAE) that compresses volatility surfaces across maturities and strikes into a 10-dimensional latent representation. We feed these latent variables, combined with option-specific inputs such as strike and maturity, into a multilayer perceptron to predict option prices. Our model is trained in stages: first to train the VAE for volatility surface compression and reconstruction, then options pricing mapping, and finally fine-tune the entire network end-to-end. The trained pricer achieves high accuracy across American and Asian options, with prediction errors concentrated primarily near long maturities and at-the-money strikes, where absolute bid-ask price differences are known to be large. Our method offers an efficient and scalable approach requiring only a single neural network forward pass and naturally improve with additional data. By bridging volatility surface modeling and option pricing in a unified framework, it provides a fast and flexible alternative to traditional numerical approaches for exotic options. △ Less
Source row: 543 · abstract type: unknown