The new NAG Library nonlinear least squares solver for unconstrained and bound-constrained data fitting (calibration) problems embeds various algorithms and regularization techniques to overcome potential convergence difficulties present in simpler methods and to increase overall robustness.
In benchmark studies, the new data fitting solver was able to solve 25% more problems and a further 60% of problems faster than the solver it replaces.
The new solver stems from a collaboration with the Rutherford Appleton Laboratory – many thanks to all involved in its development.
As with all new releases, we encourage NAG Library users upgrade to the latest Mark to access to all the new content and performance improvements. NAG Library downloads are available here. The new Mark 27.1 functionality is also available in the NAG Library for Python.
If you don’t have access to the NAG Library and you’d like to try the new functionality, we offer full product trials. If you have any questions or need help, do get in touch with our Technical Support team.
With Cloud pricing, any inefficiencies manifest as both wasted time and money. Our team of Cloud HPC experts will help you find the best way to run your HPC applications on the Cloud by optimizing for best performance and lowest cost.
- Discover the right Cloud products: Using performance, price/performance and cost of solution metrics we will identify the right Cloud infrastructure options, tuned specifically to your HPC workloads. Every HPC application is unique; right-sizing the underlying hardware is key to cost optimization.
- More efficient HPC applications: By applying traditional HPC optimization techniques, as well as porting to use cloud-specific services, we will ensure that your HPC applications are running as efficiently as possible. Wasted cycles is wasted money.
- Extensive customer application HPC experience in many different industries
- Cloud architecture, HPC administration, computing architectures, HPC applications and algorithms expertise; a mix rarely found
- NAG has certifications from all three major Cloud providers and is experienced in many other platforms; on-prem and in-cloud
- Experts located in Europe and North America ensuring customer time zone project optimization
We are certain we can help you achieve higher confidence in your Cloud HPC applications’ ROI. Do get in touch for a no-obligation consultation.
Wed, Jan 27, 2021 2:00 PM - 2:40 PM GMT – REGISTER
Long gone are the days when simple plots of parallel efficiency or parallel scaling were sufficient to understand performance issues in parallel software. It is widely accepted that analysing parallel performance is difficult, in part because of the complexity of modern parallel software, which often includes a mix of thread and process parallelism as well as GPU use, but also because of the complexity of the data generated by modern performance analysis tools.
To address this challenge the POP project, of which NAG is a partner, is developing a methodology to help code developers more easily analyse trace data and gain meaningful insight into the sources of inefficiency.
This methodology is based on the use of a small set of performance metrics, which can be generated easily from trace data, and which point at various specific code performance issues, such as process or thread load imbalance, MPI data transfer, various types of serialisation, and so on.
This webinar describes two complementary sets of metrics which can be used to identify performance bottlenecks in hybrid MPI + OpenMP software.
For some time now, NAG has partnered with the Centre for Doctoral Training in Industrially Focussed Mathematical Modelling (InFoMM) of the University of Oxford offering research topics, industrial supervision and funding for Phd students. We are proud to announce that Lindon Roberts, one of the NAG sponsored and co-supervised students, was awarded the 2019 Christopher Reddick Prize for his doctoral research “Derivative-free algorithms for nonlinear optimisation problems”. He and Coralia Cartis, his academic supervisor, also won the Mathematical Programming Computation "Best Paper of the Year" Award for: A derivative-free Gauss–Newton method. Mathematical Programming Computation (2019) 11(4):631–674 (2019).
Lindon’s algorithms were implemented in the NAG Optimization Modelling Suite (delivered with the NAG Library) – discover these specific algorithms in this mini-article and technical poster. Learn more about Lindon and the Reddick Prize here.
Congratulations to Joseph Myers Hill who was awarded the NAG Prize for best performing student in the MSc Component of the EPSRC CDT in Fluid Dynamics at the University of Leeds recently. We asked Joe to tell us a bit about his studies:
“I studied laminar fluid mixing where we use an iterative process of stretching and folding to stir fluids together, rather than relying on turbulence. This is modelled using area-preserving dynamical systems, of which we're interested in their long-term behaviour (ergodic theory and recurrence statistics). Crucially these systems are chaotic, so minute changes to a model can snowball into drastically different mixing behaviour. This also puts limits on how complex we can make the models, at least when using analytical methods. Recent work has involved testing this limit, trying to incorporate flow phenomena such as non-monotonic velocity profiles and proving that we still approach a mixed state. While this is an analytical approach, I rely heavily on numerical methods to build intuition, test hypotheses, and handle symbolic computation. Future work includes a quantitative study of recurrence in these systems, drawing together ideas from extreme value theory and chaotic dynamics to assess models which could aid in the design and improved operation of real mixing devices.”
Learn more about the EPSRC Centre for Doctoral Training in Fluid Dynamics