Carolyn Elizabeth Phelan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Carolyn Elizabeth Phelan has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Daniele Marazzina
Dipartimento di Matematica, Politecnico di Milano, Milano, Italy
6 shared publications - Guido Germano
Department of Computer Science, University College London, London, UK
6 shared publications - Paolo Barucca
University College London
4 shared publications - Kentaro Hoshisashi
Computer Science Department, University College London, London, United Kingdom
3 shared publications Gianluca Fusai
2 shared publications- Gianluca Fusai
Dipartimento di Studi per l’Economia e l’Impresa, Università del Piemonte Orientale, 28100 Novara, Italy
2 shared publications
Latest Publications
- Open Access
- 4 May 2026
- Published by American Physical Society (APS) in Physical Review E
- Vol. 113 (5) , 055301
Physics-informed neural networks for solving derivative-constrained partial differential equations
Physics-informed neural networks (PINNs) recast partial differential equations (PDEs) solving as an optimization problem in function space by minimizing a residual-based objective, yet many applications require additional derivative-based relations that are just as fundamental as the governing equations. In this paper, we present derivative-constrained PINNs (DC-PINNs), a general framework that treats constrained PDE solving as an optimization guided by a minimum objective function criterion where the physics resides in the minimum principle. DC-PINNs embed general nonlinear constraints on states and derivatives, e.g., bounds, monotonicity, convexity, incompressibility, computed efficiently via automatic differentiation, and they employ self-adaptive loss balancing to tune the influence of each objective, reducing reliance on manual hyperparameters and problem-specific architectures. DC-PINNs consistently reduce constraint violations and improve physical fidelity versus baseline PINN variants, representative hard-constraint formulations on benchmarks, including heat diffusion with bounds, financial volatilities with arbitrage-free, and fluid flow with vortices shed. Explicitly encoding derivative constraints stabilizes training and steers optimization toward physically admissible minima even when the PDE residual alone is small, providing reliable solutions of constrained PDEs grounded in energy minimum principles.
- 4 May 2026
- Published by American Physical Society (APS) in Physical Review E
- Vol. 113 (5) , 055301
Open AccessPhysics-informed neural networks for solving derivative-constrained partial differential equations
Physics-informed neural networks (PINNs) recast partial differential equations (PDEs) solving as an optimization problem in function space by minimizing a residual-based objective, yet many applications require additional derivative-based relations that are just as fundamental as the governing equations. In this paper, we present derivative-constrained PINNs (DC-PINNs), a general framework that treats constrained PDE solving as an optimization guided by a minimum objective function criterion where the physics resides in the minimum principle. DC-PINNs embed general nonlinear constraints on states and derivatives, e.g., bounds, monotonicity, convexity, incompressibility, computed efficiently via automatic differentiation, and they employ self-adaptive loss balancing to tune the influence of each objective, reducing reliance on manual hyperparameters and problem-specific architectures. DC-PINNs consistently reduce constraint violations and improve physical fidelity versus baseline PINN variants, representative hard-constraint formulations on benchmarks, including heat diffusion with bounds, financial volatilities with arbitrage-free, and fluid flow with vortices shed. Explicitly encoding derivative constraints stabilizes training and steers optimization toward physically admissible minima even when the PDE residual alone is small, providing reliable solutions of constrained PDEs grounded in energy minimum principles.
- Open Access
- 14 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 456-464
Probability‑Density‑Consistent Physics-Informed Neural Networks for Stochastic Local Volatility Model Calibration
- 14 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 456-464
Open AccessProbability‑Density‑Consistent Physics-Informed Neural Networks for Stochastic Local Volatility Model Calibration
- Open Access
- 14 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 847-855
Whack-a-mole Online Learning: Physics-Informed Neural Network for Intraday Implied Volatility Surface
Calibrating the time-dependent Implied Volatility Surface (IVS) using sparse market data is an essential challenge in computational finance, particularly for real-time applications. This task requires not only fitting market data but also satisfying a specified partial differential equation (PDE) and no-arbitrage conditions modelled by differential inequalities. This paper proposes a novel Physics-Informed Neural Networks (PINNs) approach called Whack-a-mole Online Learning (WamOL) to address this multi-objective optimisation problem. WamOL integrates self-adaptive and auto-balancing processes for each loss term, efficiently reweighting objective functions to ensure smooth surface fitting while adhering to PDE and no-arbitrage constraints and updating for intraday predictions. In our experiments, WamOL demonstrates superior performance in calibrating intraday IVS from uneven and sparse market data, effectively capturing the dynamic evolution of option prices and associated risk profiles. This approach offers an efficient solution for intraday IVS calibration, extending PINNs applications and providing a method for real-time financial modelling.
- 14 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 847-855
Open AccessWhack-a-mole Online Learning: Physics-Informed Neural Network for Intraday Implied Volatility Surface
Calibrating the time-dependent Implied Volatility Surface (IVS) using sparse market data is an essential challenge in computational finance, particularly for real-time applications. This task requires not only fitting market data but also satisfying a specified partial differential equation (PDE) and no-arbitrage conditions modelled by differential inequalities. This paper proposes a novel Physics-Informed Neural Networks (PINNs) approach called Whack-a-mole Online Learning (WamOL) to address this multi-objective optimisation problem. WamOL integrates self-adaptive and auto-balancing processes for each loss term, efficiently reweighting objective functions to ensure smooth surface fitting while adhering to PDE and no-arbitrage constraints and updating for intraday predictions. In our experiments, WamOL demonstrates superior performance in calibrating intraday IVS from uneven and sparse market data, effectively capturing the dynamic evolution of option prices and associated risk profiles. This approach offers an efficient solution for intraday IVS calibration, extending PINNs applications and providing a method for real-time financial modelling.
- Open Access
Whack-a-mole Online Learning: Physics-Informed Neural Network for Intraday Implied Volatility Surface
Calibrating the time-dependent Implied Volatility Surface (IVS) using sparse market data is an essential challenge in computational finance, particularly for real-time applications. This task requires not only fitting market data but also satisfying a specified partial differential equation (PDE) and no-arbitrage conditions modelled by differential inequalities. This paper proposes a novel Physics-Informed Neural Networks (PINNs) approach called Whack-a-mole Online Learning (WamOL) to address this multi-objective optimisation problem. WamOL integrates self-adaptive and auto-balancing processes for each loss term, efficiently reweighting objective functions to ensure smooth surface fitting while adhering to PDE and no-arbitrage constraints and updating for intraday predictions. In our experiments, WamOL demonstrates superior performance in calibrating intraday IVS from uneven and sparse market data, effectively capturing the dynamic evolution of option prices and associated risk profiles. This approach offers an efficient solution for intraday IVS calibration, extending PINNs applications and providing a method for real-time financial modelling.
Open AccessWhack-a-mole Online Learning: Physics-Informed Neural Network for Intraday Implied Volatility Surface
Calibrating the time-dependent Implied Volatility Surface (IVS) using sparse market data is an essential challenge in computational finance, particularly for real-time applications. This task requires not only fitting market data but also satisfying a specified partial differential equation (PDE) and no-arbitrage conditions modelled by differential inequalities. This paper proposes a novel Physics-Informed Neural Networks (PINNs) approach called Whack-a-mole Online Learning (WamOL) to address this multi-objective optimisation problem. WamOL integrates self-adaptive and auto-balancing processes for each loss term, efficiently reweighting objective functions to ensure smooth surface fitting while adhering to PDE and no-arbitrage constraints and updating for intraday predictions. In our experiments, WamOL demonstrates superior performance in calibrating intraday IVS from uneven and sparse market data, effectively capturing the dynamic evolution of option prices and associated risk profiles. This approach offers an efficient solution for intraday IVS calibration, extending PINNs applications and providing a method for real-time financial modelling.
- Open Access
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
The Structure and Impact of Fees on Investor and Manager Returns
Fees in fund management have attracted considerable interest over the years, both in academic literature and the financial press. Studies have shown that despite diminishing returns from hedge funds, fees have stayed high, with indications that hedge funds returns only outperform passive investments if gross returns are considered. However, investors have continued to use hedge funds and the amount of assets under management has risen. This may be related to the lack of transparency in the hedge fund industry which hampers the calculation of the realised fees paid. Existing papers have used various techniques to calculate the gross and net returns of a number of funds in the market over particular investment periods. These studies are highly informative. However they are, by design, limited to the fee arrangements of the funds in the studies and the market conditions over the relevant investment period. In this paper we take an alternate experimental approach. We have modelled fees over a number of different market conditions with a variety of fee arrangements. This allows the effect of fees on the investment to be explored under tightly controlled conditions, thus clarifying the way these mechanisms impact investors under different scenarios and time horizons. In addition the paper pays particular attention to the effect of fee frequency combined with market volatility, which is an area that has previously not been explored in detail. We show that volatile returns, coupled with the frequent application of fees, leads to very high erosion of investor capital. Furthermore, this remains the case even in the presences of hurdle rates and high-water-marks which are intended to mitigate this effect.
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe Structure and Impact of Fees on Investor and Manager Returns
Fees in fund management have attracted considerable interest over the years, both in academic literature and the financial press. Studies have shown that despite diminishing returns from hedge funds, fees have stayed high, with indications that hedge funds returns only outperform passive investments if gross returns are considered. However, investors have continued to use hedge funds and the amount of assets under management has risen. This may be related to the lack of transparency in the hedge fund industry which hampers the calculation of the realised fees paid. Existing papers have used various techniques to calculate the gross and net returns of a number of funds in the market over particular investment periods. These studies are highly informative. However they are, by design, limited to the fee arrangements of the funds in the studies and the market conditions over the relevant investment period. In this paper we take an alternate experimental approach. We have modelled fees over a number of different market conditions with a variety of fee arrangements. This allows the effect of fees on the investment to be explored under tightly controlled conditions, thus clarifying the way these mechanisms impact investors under different scenarios and time horizons. In addition the paper pays particular attention to the effect of fee frequency combined with market volatility, which is an area that has previously not been explored in detail. We show that volatile returns, coupled with the frequent application of fees, leads to very high erosion of investor capital. Furthermore, this remains the case even in the presences of hurdle rates and high-water-marks which are intended to mitigate this effect.
- Open Access
- 19 February 2020
- Published by Taylor & Francis in Quantitative Finance
- Vol. 20 (6) , 899-918
Pricing methods for α-quantile and perpetual early exercise options based on Spitzer identities
We present new numerical schemes for pricing perpetual Bermudan and American options as well as α-quantile options. This includes a new direct calculation of the optimal exercise boundary for early-exercise options. Our approach is based on the Spitzer identities for general Lévy processes and on the Wiener–Hopf method. Our direct calculation of the price of α-quantile options combines for the first time the Dassios–Port–Wendel identity and the Spitzer identities for the extrema of processes. Our results show that the new pricing methods provide excellent error convergence with respect to computational time when implemented with a range of Lévy processes.
- 19 February 2020
- Published by Taylor & Francis in Quantitative Finance
- Vol. 20 (6) , 899-918
Open AccessPricing methods for α-quantile and perpetual early exercise options based on Spitzer identities
We present new numerical schemes for pricing perpetual Bermudan and American options as well as α-quantile options. This includes a new direct calculation of the optimal exercise boundary for early-exercise options. Our approach is based on the Spitzer identities for general Lévy processes and on the Wiener–Hopf method. Our direct calculation of the price of α-quantile options combines for the first time the Dassios–Port–Wendel identity and the Spitzer identities for the extrema of processes. Our results show that the new pricing methods provide excellent error convergence with respect to computational time when implemented with a range of Lévy processes.
- Open Access
- 26 April 2019
- Published by Elsevier in SSRN Electronic Journal
Pricing Methods for Alpha-Quantile and Perpetual Early Exercise Options Based on Spitzer Identities
We present new numerical schemes for pricing perpetual Bermudan and American options as well as α-quantile options. This includes a new direct calculation of the optimal exercise barrier for early-exercise options. Our approach is based on the Spitzer identities for general Lèvy processes and on the Wiener-Hopf method. Our direct calculation of the price of alpha-quantile options combines for the first time the Dassios-Port-Wendel identity and the Spitzer identities for the extrema of processes. Our results show that the new pricing methods provide excellent error convergence with respect to computational time when implemented with a range of Lèvy processes.
- 26 April 2019
- Published by Elsevier in SSRN Electronic Journal
Open AccessPricing Methods for Alpha-Quantile and Perpetual Early Exercise Options Based on Spitzer Identities
We present new numerical schemes for pricing perpetual Bermudan and American options as well as α-quantile options. This includes a new direct calculation of the optimal exercise barrier for early-exercise options. Our approach is based on the Spitzer identities for general Lèvy processes and on the Wiener-Hopf method. Our direct calculation of the price of alpha-quantile options combines for the first time the Dassios-Port-Wendel identity and the Spitzer identities for the extrema of processes. Our results show that the new pricing methods provide excellent error convergence with respect to computational time when implemented with a range of Lèvy processes.
- Open Access
- 1 November 2018
- Published by Elsevier in European Journal of Operational Research
- Vol. 271 (1) , 210-223
Fluctuation identities with continuous monitoring and their application to the pricing of barrier options
- 1 November 2018
- Published by Elsevier in European Journal of Operational Research
- Vol. 271 (1) , 210-223
Open AccessFluctuation identities with continuous monitoring and their application to the pricing of barrier options
- 14 May 2018
- Published by Springer Nature in Annals of Operations Research
- Vol. 282 (1-2) , 273-298
Hilbert transform, spectral filters and option pricing
- 14 May 2018
- Published by Springer Nature in Annals of Operations Research
- Vol. 282 (1-2) , 273-298
Hilbert transform, spectral filters and option pricing
- Open Access
Fluctuation identities with continuous monitoring and their application to price barrier options
We present a numerical scheme to calculate fluctuation identities for exponential L\'evy processes in the continuous monitoring case. This includes the Spitzer identities for touching a single upper or lower barrier, and the more difficult case of the two-barriers exit problem. These identities are given in the Fourier-Laplace domain and require numerical inverse transforms. Thus we cover a gap in the literature that has mainly studied the discrete monitoring case; indeed, there are no existing numerical methods that deal with the continuous case. As a motivating application we price continuously monitored barrier options with the underlying asset modelled by an exponential L\'evy process. We perform a detailed error analysis of the method and develop error bounds to show how the performance is limited by the truncation error of the sinc-based fast Hilbert transform used for the Wiener-Hopf factorisation. By comparing the results for our new technique with those for the discretely monitored case (which is in the Fourier-$z$ domain) as the monitoring time step approaches zero, we show that the error convergence with continuous monitoring represents a limit for the discretely monitored scheme.
Open AccessFluctuation identities with continuous monitoring and their application to price barrier options
We present a numerical scheme to calculate fluctuation identities for exponential L\'evy processes in the continuous monitoring case. This includes the Spitzer identities for touching a single upper or lower barrier, and the more difficult case of the two-barriers exit problem. These identities are given in the Fourier-Laplace domain and require numerical inverse transforms. Thus we cover a gap in the literature that has mainly studied the discrete monitoring case; indeed, there are no existing numerical methods that deal with the continuous case. As a motivating application we price continuously monitored barrier options with the underlying asset modelled by an exponential L\'evy process. We perform a detailed error analysis of the method and develop error bounds to show how the performance is limited by the truncation error of the sinc-based fast Hilbert transform used for the Wiener-Hopf factorisation. By comparing the results for our new technique with those for the discretely monitored case (which is in the Fourier-$z$ domain) as the monitoring time step approaches zero, we show that the error convergence with continuous monitoring represents a limit for the discretely monitored scheme.