European Commission logo
English English
CORDIS - EU research results
Content archived on 2024-06-18

Training in Modern Quantitative Methods and High-Performance Computing for Finance

Final Report Summary - HPCFINANCE (Training in Modern Quantitative Methods and High-Performance Computing for Finance)

HPCFinance stands at the fertile crossroads Financial Engineering and High Performance Computing. The project has developed and integrated the realistic and effective financial models for successful risk management and pricing of derivatives on different HPC platforms. The availability of substantial computing resources has allowed the project to use and develop theories and methods more flexibly, and to relax unrealistic assumptions so far applied to reduce the computation burden. The project has had three interconnected areas: methodologies and modeling; quantitative risk management and derivative pricing; and HPC engineering. The multidisciplinary program has consisted of modular work streams in these fields, helping the industry to apply the newest ideas and the most up-to-date methods.

HPCFinance ITN recruited under employment contract 14 Research Projects (RPs) providing training to 12 ESRs and 2 ERs. Each fellow has had own research project ( All the ESRs have been seconded to network’s partner organisations at least for two months and ERs for one month. All fellows have also collaborated with numerous external private sector organizations and universities.

The project has been developing and implementing state-of-the-art models and methods for derivative pricing and risk management. A number of journal papers are published and more than 50 conference presentations have been provided sharing the results from HPCFinance research projects (visit for more information). Most important publications as an outcome of HPCFinance research include the following papers:

Dempster, M.A.H. Kloppers, D., Medova, E., Osmolovskiy, I. & Ustinov, P. (2016), “Life cycle goal achievement or portfolio volatility reduction?”. Journal of Portfolio Management, Vol. 42, No.2 pp. 99-117.
This paper is concerned with the use of currently available technology to provide individuals, financial advisors and pension fund financial planners with detailed prospective financial plans tailored to an individual's financial goals and obligations. The performance of the adaptive dynamic goal-based portfolio strategy is found to be far superior to all the industry’s Markowitz-based approaches. These empirical results should put paid to the commonly held view amongst finance professionals that the extra complexity of holistic dynamic stochastic models is not worth the marginal extra value obtained from their employment.

Yang, H. and J. Kanniainen (2016), “Jump and Volatility Dynamics for the S&P500: Evidence for Infinite-Activity Jumps with Non-Affine Volatility Dynamics from Stock and Option Markets”, forthcoming in the Review of Finance.
This paper uses extensive empirical data sets to study how infinite-activity Variance Gamma and Normal Inverse Gaussian (NIG) jumps with affine and non-affine volatility dynamics improve goodness of fit and option pricing performance. The article provides clear evidence that a parsimonious non-affine model with NIG return jumps and a linear variance specification is particularly competitive, even during the recent crisis. The research results are important in derivative pricing and risk management.

Pelsser, A and Salahnejhad Ghalehjooghi, A (2016) “Time-consistent actuarial valuations”, Insurance: Mathematics and Economics, Vol. 66, pp. 97-112.
This paper investigates the continuous-time limits of well-known actuarial premium principles when backward iteration procedures are applied to an insurance risk process in the form of a diffusion process and a jump process in order to capture the heavy tailed nature of insurance liabilities. In the case of the diffusion process, the one-period time-consistent Variance premium principle converges to the non-linear exponential indifference price and the Standard-Deviation and the Cost-of-Capital principle converge to the same price limit. Adding the jump risk gives a more realistic picture of the price. The VaR operator fails to capture the jump risk for small jump probabilities, and the time-consistent price depends on the distribution of the premium jump. The results are useful in actuarial valuations.

Kozikowski, G., Papamanousakis, G., Yang, J. (2015), "Potential future exposure, modelling and accelerating on GPI and FPGA", WHPCF'15 Proceedings of the 8th Workshop on High Perfomance Computational Finance
This paper proposes a PFE model that fits specific business requirements, as well as a GPU and a FPGA implementation of such model. The FPGA implementation has been optimised in terms of the performance to support a fully pipelined design. Experimental results show that the GPU implementation can achieve up to 25 times speedup over CPU solution, and the FPGA implementation can achieve up to 120 times speedup.

Kozikowski, G. & Kubica, B. (2014), "Parallel Approach to Monte Carlo Simulation for Option Price Sensitivities Using the Adjoint and Interval Analysis", Parallel Processing and Applied Mathematics, Lecture Notes in Computer Science, pp. 600-612.
This paper concerns a new approach to evaluation of Option Price sensitivities using the Monte Carlo simulation, based on the parallel GPU architecture and Automatic Differentiation methods. Considerations are based on two implementations of the algorithm – the sequential and parallel ones. For efficient differentiation, the Adjoint method is employed. Computational experiments include analysis of performance, uncertainty error and rounding error and consider Black-Scholes and Heston models. The results can be exploited e.g. in delta-hedging strategies.

Martino, L., H. Yang, D. Luengo, J. Kanniainen, J. Corander (2015), “A Fast Universal Self-tuned Sampler within Gibbs sampling”, Digital Signal Processing, 47, pp. 68-83.
In this work, we present a simple, self-tuned and extremely efficient MCMC algorithm which produces virtually independent samples from these univariate target densities. Numerical experiments, on several synthetic data sets and a high-dimensional financial signal processing problem, show its good performance in terms of speed and estimation accuracy. The results can be used to estimate volatility models in a more robust way in the academia and the financial industry.

Hu, J. and J. Kanniainen (2015), “Asymptotic expansion of European options with mean-reverting stochastic volatility dynamics”, Finance Research Letters, 14, pp. 1-10.
We develop methods for pricing European options under general mean-reverting stochastic volatility dynamics, which can be used with both affine and non-affine volatility models. The analytic approximation is more generally applicable than the fast Fourier transform, because it does not rely on the existence of a characteristic function. We numerically demonstrate our approach with the Heston, 3/2, and continuous-time GARCH models. These results are important in risk management and derivative pricing.

Kanniainen, J., Lin, B. & Yang, H. (2014), “Estimating and using GARCH models with VIX data for option valuation”, Journal of Banking and Finance, Vol. 43, pp. 200-211.
This paper uses information on VIX to improve the empirical performance of GARCH models for pricing options on the S&P 500. The models’ performance can clearly be improved by extracting daily spot volatilities from the series of VIX rather than by using the series of the underlying’s returns. Moreover, , a joint MLE with returns and VIX improves option pricing performance. Finally, consistently with the existing research, this paper finds that non-affine models clearly outperform affine models. These results are important in risk management and derivative pricing.

Kozikowski, G. & Kubica, B. (2013), "Interval Arithmetic and Automatic Differentiation on GPU Using OpenCL", Applied Parallel and Scientific Computing, Lecture Notes in Computer Science, Vol. 7782, pp. 489-503.
This paper investigates efficient and powerful approach to the Gradient and the Hessian evaluation for complex functions. Computational experiments include analysis of performance and are studied on the generated test functions with a given complexity.

Baldi, P. and Pisani, C. (2013) "Simple Simulation Schemes for CIR and Wishart Processes", International Journal of Theoretical and Applied Finance, Vol. 16, No, 8.
This paper develops some simple simulation algorithms for CIR and Wishart processes. We investigate rigorously the square of a matrix valued Ornstein–Uhlenbeck process, the main idea being to split the generator and to reduce the problem to the simulation of the square of a matrix valued Ornstein–Uhlenbeck process to be added to a deterministic process. In this way, we provide a weak second-order scheme that requires only the simulation of i.i.d. Gaussian r.v.'s and simple matrix manipulations. These results are important in risk management activities.

Project website:
Contact details:
Juho Kanniainen, Prof. (HPCFinance Project Coordinator)
Tampere University of Technology
Tel: +358 40 707 4532
Email: juho.kanniainen (at)