Technical Posters Archive - nAG

Portfolio Optimization using SOCP in the nAG Library

Lenny Pitt — Fri, 10 Nov 2023 11:42:44 +0000

Second-order cone programming (SOCP) is convex optimization which extends linear programming (LP) with second-order (Lorentz or the ice cream) cones.

The post Portfolio Optimization using SOCP in the nAG Library appeared first on nAG.

nZetta Derivatives Pricing Toolkit for PDE

Lenny Pitt — Tue, 26 Sep 2023 10:12:56 +0000

Designed from the outset for modern hardware, the nZetta Toolkit achieves a
10x performance improvement over reference 2D-PDE calculations on real finance problems.

The post nZetta Derivatives Pricing Toolkit for PDE appeared first on nAG.

nZetta Derivatives Pricing Toolkit for Monte Carlo

Lenny Pitt — Tue, 26 Sep 2023 10:05:02 +0000

With Monte Carlo, it is crucial to leverage the full potential of hardware parallelism to achieve optimal performance and accuracy.

The post nZetta Derivatives Pricing Toolkit for Monte Carlo appeared first on nAG.

Faster Risk Calculation: Next Generation dco/c++

Johannes Lotz — Mon, 31 Jul 2023 17:50:07 +0000

Trading Desks can hedge auspiciously and gain a competitive edge in the market as our mathematical algorithm technology, dco/c++, gives them rich, cheap and accurate intra-day risk. This fast and accurate risk data at lower cost means more profits for traders and for the business.

The post Faster Risk Calculation: Next Generation dco/c++ appeared first on nAG.

Nearest Correlation Algorithms in the nAG Library

NAG — Wed, 21 Jun 2023 10:32:05 +0000

An approximate correlation matrix is one that is not positive semidefinite.
In this document we consider an application from finance.

The post Nearest Correlation Algorithms in the nAG Library appeared first on nAG.

CVA in the Cloud

NAG — Wed, 21 Jun 2023 10:25:23 +0000

NAG has developed a CVA demonstration code to show how the NAG Library and the Algorithmic Differentiation (AD) tool dco/c++ can be combined with Origami to solve large scale CVA computations.

The post CVA in the Cloud appeared first on nAG.

Fast Implied Volatilities using Chebyshev Interpolation

NAG — Wed, 21 Jun 2023 10:20:46 +0000

An important problem in finance is to compute the implied volatility. Typically volatilities are computed for large vectors of input data.

The post Fast Implied Volatilities using Chebyshev Interpolation appeared first on nAG.

CVA at Scale with Adjoint Sensitivities

NAG — Wed, 21 Jun 2023 10:17:31 +0000

Calculating XVA in a timely manner poses a big performance challenge
for financial institutions.

The post CVA at Scale with Adjoint Sensitivities appeared first on nAG.

Performance Analysis of a GS2 Plasma Turbulence Code

NAG — Wed, 21 Jun 2023 10:07:56 +0000

GS2 [1] is an open source gyrokinetic simulation code used to study turbulence in plasma, one application is for fusion experiments. It is a gyrokinetic flux tube initial value and eigenvalue solv-er and is written in Fortran and parallelised with MPI.

The post Performance Analysis of a GS2 Plasma Turbulence Code appeared first on nAG.

Performance and Usage Improvements for Multiple OpenFOAM Customers: HPC and Cloud

NAG — Wed, 21 Jun 2023 10:02:00 +0000

This poster describes work performed on OpenFOAM focussing on
performance as well as OpenFOAM in the Cloud.

The post Performance and Usage Improvements for Multiple OpenFOAM Customers: HPC and Cloud appeared first on nAG.

Adjoint Automatic Differentiation of a GPU Accelerated Application

Lenny Pitt — Wed, 21 Jun 2023 09:51:49 +0000

Adjoints are sophisticated numerical techniques for computing a large
number of gradients quickly. To compute an adjoint, your computer
program must be run backwards.

The post Adjoint Automatic Differentiation of a GPU Accelerated Application appeared first on nAG.

High Performance Tape–Free Adjoint AD for C++11

NAG — Wed, 21 Jun 2023 09:00:09 +0000

The C++ language is so complex that no AD compilers can handle it.
To get an adjoint, we must write it by hand or use an operator overloading
AAD tool.

The post High Performance Tape–Free Adjoint AD for C++11 appeared first on nAG.