What is Premia

Premia is a software designed for option pricing, hedging and financial model calibration. It is provided with it's C/C++ source code and an extensive scientific documentation.

Efficient computations of prices and hedges for derivative products are major issues for financial institutions. The development of increasingly complex financial products requires the use of advanced stochastic and numerical analysis techniques. A consortium of banks have been using Premia since it's beginning in 1999 and have brought important contributions to the project.

Premia is developed by the MathRisk team which gathers research scientists from INRIA (the French national institute for research in computer science and control), CERMICS / Ecole des Ponts (ENPC) – laboratory on applied mathematics and computing.

Major features of Premia :

  • Pricing of Interest Rate Derivatives
  • Pricing of Credit Risk Derivatives
  • Pricing and Hedging of Equity Derivatives in Black-Scholes and Heston models
  • Pricing and Hedging of Equity Derivatives in Jump models
  • Calibration in Jump models
  • Neural Network and Machine Learning for High-Dimensional American Option Pricing
  • Chebyshev Interpolation for Parametric Option Pricing
  • Deep Hedging
  • Computation of XVA
  • Numerical methods for rough stochastic volatility models
  • GPUs for Option Pricing
  • Pricing of Variables Annuities

Overview

Scientific Background

PREMIA is a digital platform for quantitative finance developed as part of the MathRisk project at Inria Paris and ENPC/Cermics.
PREMIA provides efficient implementations of numerical methods for pricing financial options on equities, interest rates, inflation, credit, energy, and insurance products. The following algorithms are implemented: deterministic methods (finite differences, finite elements for partial differential equations, wavelets, Galerkin methods, and sparse grids); stochastic algorithms (Monte Carlo, quantization, and methods based on Malliavin calculus); tree-based methods; approximation methods (Laplace transforms, fast Fourier transforms, Wiener-Hopf factorisation). Python-based research includes neural-network pricing, deep optimal stopping and deep hedging.

Technological Background and Characteristics of the PREMIA Software

PREMIA consists of a library for describing financial products, models, and pricing methods; a collection of C/C++ pricing routines; and a scientific documentation system in PDF and HTML formats covering the implemented algorithms. Machine learning algorithms are primarily developed in Python.
The PREMIA software is built around a core written in C, which provides an object-oriented architecture designed to represent the four fundamental components used to define a derivative pricing problem: the underlying asset type, the model, the option type, and the associated pricing method.
In the context of scientific development of machine learning algorithms, part of the PREMIA software is written in Python and uses standard scientific computing libraries (NumPy, SciPy) and machine learning libraries (PyTorch, TensorFlow, etc.). These modules are independent of the core and can be executed in a minimal Python environment, as described in the documentation for each module.

Models and Derivative Instruments

Equity and exotic derivatives. PREMIA covers European, American and Bermudan-style vanilla and exotic options, including multi-asset and path-dependent payoffs. Core modelling frameworks include multi-dimensional Black–Scholes, Dupire local volatility, CEV, Heston and other stochastic-volatility specifications, together with jump-diffusion and exponential Lévy models such as Merton, Kou, variance gamma and normal inverse Gaussian. The scientific catalogue also considers SABR, stochastic-local volatility, Bergomi, rough-volatility and Bates-type models.

Interest-rate, inflation and hybrid derivatives. PREMIA covers a broad range of interest-rate derivatives and structured fixed-income products. Models include Vasicek, Hull–White, CIR/CIR++, multifactor and affine term-structure specifications, Heath–Jarrow–Morton and LIBOR market models, including jump and stochastic-volatility extensions. The catalogue also covers inflation-indexed derivatives and hybrid contracts involving interest-rate, inflation and foreign-exchange risk.

Credit derivatives and counterparty risk. Applications include single-name and basket credit default swaps (CDS), collateralised debt obligations (CDOs) and portfolio-credit structures under reduced-form intensity, copula and portfolio-loss models. Related research covers exposure estimation, credit valuation adjustment (CVA) and broader XVA methodologies, including GPU-accelerated numerical methods.

Commodities, volatility and insurance. PREMIA addresses energy swing options and commodity spread options, with mean-reverting and jump-driven price dynamics. The scientific catalogue also covers variance and volatility swaps and options on realised variance. Insurance applications include equity-linked guarantees and variable annuities with guaranteed minimum withdrawal and guaranteed lifetime withdrawal benefits (GMWB and GLWB), including stochastic-volatility and stochastic-interest-rate settings.

Consortium using Premia

Premia is developed in interaction with a consortium of financial institutions presently composed of Crédit Agricole Corporate and Investment Bank, Natixis.

The participants of the consortium finance the development of Premia (by contributing to the salaries of expert engineers hired by the MathRisk project every year to develop the software) and help to determine the directions in which the project evolves.

Financial institutions interested in joining the consortium may contact  premia@inria.fr.

Every year, during a "delivery meeting", a new version of Premia is presented to the consortium by the members of the MathRisk project working on the software. This presentation is followed by a discussion on the features to be incorporated in the next release.

In addition, between delivery meetings, MathRisk project members meet individual consortium participants to further clarify their needs and interests.

After the release of each new version of Premia, two year older versions become available on this web site and can be freely downloaded for academic purposes.

The consortium is composed of the following financial institutions :

Current Premia team

Development leader

  • Ludovic GOUDENEGE, CNRS, LaMME, Univ. Paris-Saclay, France
  • Jérôme LELONG, Grenoble INP - UGA, France

Permanent contributors

  • Aurélien ALFONSI, Ecole des Ponts (ENPC), France
  • Julien GUYON, Ecole des Ponts (ENPC), France
  • Benjamin JOURDAIN, Ecole des Ponts (ENPC), France
  • Ahmed KEBAIER, LaMME, Univ. Evry Paris-Saclay, France
  • Xiao WEI, CUFE Beijing, China

Administrative assistant

  • Diana MARINO DUARTE,, INRIA, France

Scientific leader

  • Antonino ZANETTE, Udine University, Italy

Steering committee

  • Aurélien ALFONSI, Ecole des Ponts (ENPC), France
  • Agnès SULEM, INRIA, France

Scientific partners

  • Vlad BALLY, Gustave Eiffel University, France
  • Arturo KOHATSU-HIGA, Ritsumeikan University , Japan
  • Nicolas PRIVAULT, NTU, Singapore

Contact

For any questions about technical aspects of the Premia Software package please contact: premia@inria.fr

For any questions about a presentation of

  • Premia
  • the MathRisk project

please contact:

Agnès Sulem

INRIA
Centre de recherche Inria de Paris
48, rue Barrault
CS 61534
75647 PARIS CEDEX
e-mail: agnes.sulem@inria.fr

Aurélien Alfonsi

CERMICS / Ecole des Ponts (ENPC)
6 et 8 avenue Blaise Pascal
Cité Descartes
77455 Champs-sur-Marne
Marne -la-Vallée cedex 2
email: aurelien.alfonsi@enpc.fr



Research project done with CERMICS / Ecole des Ponts (ENPC)

Premia    Premia