Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Generating Volatility Surfaces using Variational Autoencoders

Immo Barasciutti, Michael LU and Fossum, Axel LU (2024) In Master's Theses in Mathematical Sciences FMAM05 20241
Mathematics (Faculty of Engineering)
Abstract
This study presents an in-depth exploration into the utilization of Variational Autoencoders (VAEs) for modeling and completing implied volatility surfaces within the context of the index equities market, a crucial aspect of option pricing. Moreover, our study examines the predictive capabilities of neural networks concerning fluctuations in spot prices, with a specialized spot model calibrated to forecast changes in volatility surfaces based on spot price dynamics. Through comprehensive data processing and structuring of VAEs we created a model capable of generating accurate and nearly arbitrage-free volatility surfaces from as little as 10 points of information. This model also proved proficiency in generating volatility surfaces for... (More)
This study presents an in-depth exploration into the utilization of Variational Autoencoders (VAEs) for modeling and completing implied volatility surfaces within the context of the index equities market, a crucial aspect of option pricing. Moreover, our study examines the predictive capabilities of neural networks concerning fluctuations in spot prices, with a specialized spot model calibrated to forecast changes in volatility surfaces based on spot price dynamics. Through comprehensive data processing and structuring of VAEs we created a model capable of generating accurate and nearly arbitrage-free volatility surfaces from as little as 10 points of information. This model also proved proficiency in generating volatility surfaces for previously unseen underlying assets. Applying changes in spot price as a conditional variable we successfully created a powerful risk management tool capable of forecasting volatility surfaces for various future scenarios.

Although our model can be improved upon, our findings underscore the robustness and generalizability of VAEs, showcasing their potential for broader application across various financial instruments and markets. (Less)
Popular Abstract (Swedish)
Generering av volatilitetsytor är en väsentlig del i prissättning av optioner. Dagens modeller är mycket komplexa och saknar egenskapen att kunna representera alla möjliga ytor, därför ses neurala nätverk som ett alternativ. I denna rapport undersöker vi användbarheten av just Variational Autoencoders på aktiemarknaden.
Please use this url to cite or link to this publication:
author
Immo Barasciutti, Michael LU and Fossum, Axel LU
supervisor
organization
course
FMAM05 20241
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Variational Autoencoder (VAE), static arbitrage, implied volatility, volatility surface, index equity options, neural network, machine learning, quantitative finance
publication/series
Master's Theses in Mathematical Sciences
report number
LUTFMA-3534-2024
ISSN
1404-6342
other publication id
2024:E22
language
English
id
9153880
date added to LUP
2024-06-28 16:05:17
date last changed
2024-06-28 16:05:17
@misc{9153880,
  abstract     = {{This study presents an in-depth exploration into the utilization of Variational Autoencoders (VAEs) for modeling and completing implied volatility surfaces within the context of the index equities market, a crucial aspect of option pricing. Moreover, our study examines the predictive capabilities of neural networks concerning fluctuations in spot prices, with a specialized spot model calibrated to forecast changes in volatility surfaces based on spot price dynamics. Through comprehensive data processing and structuring of VAEs we created a model capable of generating accurate and nearly arbitrage-free volatility surfaces from as little as 10 points of information. This model also proved proficiency in generating volatility surfaces for previously unseen underlying assets. Applying changes in spot price as a conditional variable we successfully created a powerful risk management tool capable of forecasting volatility surfaces for various future scenarios. 

Although our model can be improved upon, our findings underscore the robustness and generalizability of VAEs, showcasing their potential for broader application across various financial instruments and markets.}},
  author       = {{Immo Barasciutti, Michael and Fossum, Axel}},
  issn         = {{1404-6342}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Master's Theses in Mathematical Sciences}},
  title        = {{Generating Volatility Surfaces using Variational Autoencoders}},
  year         = {{2024}},
}