• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 2
  • Tagged with
  • 2
  • 2
  • 2
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
1

Deep learning exotic derivatives

Geirsson, Gunnlaugur January 2021 (has links)
Monte Carlo methods in derivative pricing are computationally expensive, in particular for evaluating models partial derivatives with regard to inputs. This research proposes the use of deep learning to approximate such valuation models for highly exotic derivatives, using automatic differentiation to evaluate input sensitivities. Deep learning models are trained to approximate Phoenix Autocall valuation using a proprietary model used by Svenska Handelsbanken AB. Models are trained on large datasets of low-accuracy (10^4 simulations) Monte Carlo data, successfully learning the true model with an average error of 0.1% on validation data generated by 10^8 simulations. A specific model parametrisation is proposed for 2-day valuation only, to be recalibrated interday using transfer learning. Automatic differentiation approximates sensitivity to (normalised) underlying asset prices with a mean relative error generally below 1.6%. Overall error when predicting sensitivity to implied volatililty is found to lie within 10%-40%. Near identical results are found by finite difference as automatic differentiation in both cases. Automatic differentiation is not successful at capturing sensitivity to interday contract change in value, though errors of 8%-25% are achieved by finite difference. Model recalibration by transfer learning proves to converge over 15 times faster and with up to 14% lower relative error than training using random initialisation. The results show that deep learning models can efficiently learn Monte Carlo valuation, and that these can be quickly recalibrated by transfer learning. The deep learning model gradient computed by automatic differentiation proves a good approximation of the true model sensitivities. Future research proposals include studying optimised recalibration schedules, using training data generated by single Monte Carlo price paths, and studying additional parameters and contracts.
2

Pricing With Uncertainty : The impact of uncertainty in the valuation models ofDupire and Black&Scholes

Zetoun, Mirella January 2013 (has links)
Theaim of this master-thesis is to study the impact of uncertainty in the local-and implied volatility surfaces when pricing certain structured products suchas capital protected notes and autocalls. Due to their long maturities, limitedavailability of data and liquidity issue, the uncertainty may have a crucialimpact on the choice of valuation model. The degree of sensitivity andreliability of two different valuation models are studied. The valuation models chosen for this thesis are the local volatility model of Dupire and the implied volatility model of Black&Scholes. The two models are stress tested with varying volatilities within an uncertainty interval chosen to be the volatilities obtained from Bid and Ask market prices. The volatility surface of the Mid market prices is set as the relative reference and then successively scaled up and down to measure the uncertainty.The results indicates that the uncertainty in the chosen interval for theDupire model is of higher order than in the Black&Scholes model, i.e. thelocal volatility model is more sensitive to volatility changes. Also, the pricederived in the Black&Scholes modelis closer to the market price of the issued CPN and the Dupire price is closer tothe issued Autocall. This might be an indication of uncertainty in thecalibration method, the size of the chosen uncertainty interval or the constantextrapolation assumption.A further notice is that the prices derived from the Black&Scholes model areoverall higher than the prices from the Dupire model. Another observation ofinterest is that the uncertainty between the models is significantly greaterthan within each model itself. / Syftet med dettaexamensarbete är att studera inverkan av osäkerhet, i prissättningen av struktureradeprodukter, som uppkommer på grund av förändringar i volatilitetsytan. I dennastudie värderas olika slags autocall- och kapitalskyddade struktureradeprodukter. Strukturerade produkter har typiskt långa löptider vilket medförosäkerhet i värderingen då mängden data är begränsad och man behöver ta tillextrapolations metoder för att komplettera. En annan faktor som avgörstorleksordningen på osäkerheten är illikviditeten, vilken mäts som spreadenmellan listade Bid och Ask priset. Dessa orsaker ligger bakom intresset attstudera osäkerheten för långa löptider över alla lösenpriser och dess inverkanpå två olika värderingsmodeller.Värderingsmodellerna som används i denna studie är Dupires lokala volatilitetsmodell samt Black&Scholes implicita volatilitets modell. Dessa ställs motvarandra i en jämförelse gällande stabilitet och förmåga att fånga uppvolatilitets ändringar. Man utgår från Mid volatilitetsytan som referens ochuppmäter prisändringar i intervallet från Bid upp till Ask volatilitetsytornagenom att skala Mid ytan. Resultaten indikerar på större prisskillnader inom Dupires modell i jämförelsemot Black&Scholes. Detta kan tolkas som att Dupires modell är mer känslig isammanhanget och har en starkare förmåga att fånga upp förändringar isvansarna. Vidare notering är att priserna beräknade i Dupire är relativtbilligare än motsvarande från Black&Scholes modellen. En ytterligareobservation är att osäkerheten mellan värderingsmodellerna är av högre ordningän inom var modell för sig. Ett annat resultat visar att CPN priset beräknat iBlack&Scholes modell ligger närmast marknadspriset medans marknadsprisetför Autocallen ligger närmare Dupires. Detta kan vara en indikation påosäkerheten i kalibreringsmetoden eventuellt det valda osäkerhetsintervalletoch konstanta extrapolations antagandet.

Page generated in 0.0262 seconds