Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Option Pricing Using Forward Curves From Arbitrage Free Autoencoders

von Ekensteen Löfgren, Linus LU (2025) NEKH02 20251
Department of Economics
Abstract (Swedish)
This thesis evaluates the feasibility of arbitrage free autoencoder models for forward curve simulation using European bond data. In the model, an autoencoder is used to learn a low dimensional factor representation of forward curves. The risk neutral evolution of forward curves can be calculated by introducing a stochastic process in this low dimensional latent space. A correction factor is added to make the evolution free of arbitrage. The model is found to be well suited for encoding forward curves in a low dimensional latent space. Pricing is evaluated by investigating the difference in dynamics between the autoencoder model and the canonical Hull-White model. While difficulties in calibration are noted, the autoencoder model performs... (More)
This thesis evaluates the feasibility of arbitrage free autoencoder models for forward curve simulation using European bond data. In the model, an autoencoder is used to learn a low dimensional factor representation of forward curves. The risk neutral evolution of forward curves can be calculated by introducing a stochastic process in this low dimensional latent space. A correction factor is added to make the evolution free of arbitrage. The model is found to be well suited for encoding forward curves in a low dimensional latent space. Pricing is evaluated by investigating the difference in dynamics between the autoencoder model and the canonical Hull-White model. While difficulties in calibration are noted, the autoencoder model performs well when pricing using Monte Carlo. (Less)
Please use this url to cite or link to this publication:
author
von Ekensteen Löfgren, Linus LU
supervisor
organization
course
NEKH02 20251
year
type
M2 - Bachelor Degree
subject
language
English
id
9211407
date added to LUP
2025-09-12 09:16:33
date last changed
2025-09-12 09:16:33
@misc{9211407,
  abstract     = {{This thesis evaluates the feasibility of arbitrage free autoencoder models for forward curve simulation using European bond data. In the model, an autoencoder is used to learn a low dimensional factor representation of forward curves. The risk neutral evolution of forward curves can be calculated by introducing a stochastic process in this low dimensional latent space. A correction factor is added to make the evolution free of arbitrage. The model is found to be well suited for encoding forward curves in a low dimensional latent space. Pricing is evaluated by investigating the difference in dynamics between the autoencoder model and the canonical Hull-White model. While difficulties in calibration are noted, the autoencoder model performs well when pricing using Monte Carlo.}},
  author       = {{von Ekensteen Löfgren, Linus}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Option Pricing Using Forward Curves From Arbitrage Free Autoencoders}},
  year         = {{2025}},
}