Artelys Knitro 15.0: New Tools for Your Large-Scale Models
Artelys is pleased to announce the release of Knitro 15.0, which provides new algorithms and performance improvements to solve your large-scale optimisation problems, whether linear or non-linear, more quickly
New Algorithms
With version 15.0, Knitro expands its range of algorithms to solve your optimisation problems:
- Augmented Lagrangian Method for Non-linear Problems :
Specifically designed for non-linear optimisation models, this algorithm uses a non-linearly constrainted approach and can take advantage of second derivatives. It is particularly suitable for complex problems with degenerate constraints where the linear independence constraint qualification (LICQ) is not satisfied.
- Primal-Dual Hybrid Gradient (PDHG) Algorithm:
Developed to solve linear programming problems, this algorithm uses only first-order information to converge and limits matrix factorizations. It is ideal for very large-scale models when interior point methods encounter difficulties.
New Features
- Model Re-optimisation:
The Knitro API has been updated to allow re-optimisation of a model after modifications to quadratic and non-linear structures.
- Configurable Cut-off for Branch-and-Bound:
New parameters have been introduced to set absolute and relative cut-off levels in the Branch-and-Bound algorithm, offering better control over optimisation processes.
Improved Performance
Knitro 15.0 enhances the efficiency and robustness of existing optimisation methods:
- Faster Resolution of MILP and MINLP Problems:
The addition of new presolve operations allows for faster resolution of your mixed-integer programming models, particularly for non-linear problems (MINLP) with linear structures.
- Faster Hessian Approximation:
This version offers significant performance improvements for Hessian approximation with the limited-memory Broyden-Fletcher-Goldfarb-Shanno method, especially for large-scale problems.
- Use of Apple Accelerate Sparse Solvers Library:
Users of Mac-ARM platforms can now use Apple Accelerate’s specific solvers for sparse matrix equation systems, thereby improving performance on large-scale models.
Thanks to these improvements, Knitro 15.0 reduces solution times on continuous problems on our pooling and LP datasets by 46% and 14% respectively, compared with version 14.2. Our evaluation datasets are composed from academic and customer cases instances, with application in water networks, oil and gas, refinery processing, etc.
Knitro 15.0 is also more robust on discrete problems, with new instances solved on our MINLP convex dataset. On the latter, the primal-dual integral gap is reduced by more than 10%, demonstrating increased convergence speed and solution quality.
More Accessible Interfaces
- Python Interface Available on PyPI:
The installation process for Knitro’s Python interface has been completely revamped. It is now possible to install Knitro and its Python interface directly from PyPI, the Python package manager, greatly facilitating the use of our non-linear solver from this language.
- Java Interface Compatible with Mac ARM:
With this new version, Knitro’s Java interface is now available for users of Mac ARM platforms.
To learn more about Knitro 15.0 and its capabilities, visit the official documentation or review the detailed release notes.
New Algorithms
With version 15.0, Knitro expands its range of algorithms to solve your optimisation problems:
- Augmented Lagrangian Method for Non-linear Problems: Specifically designed for non-linear optimisation models, this algorithm uses a non-linearly constrainted approach and can take advantage of second derivatives. It is particularly suitable for complex problems with degenerate constraints where the linear independence constraint qualification (LICQ) is not satisfied.
- Primal-Dual Hybrid Gradient (PDHG) Algorithm: Developed to solve linear programming problems, this algorithm uses only first-order information to converge and limits matrix factorizations. It is ideal for very large-scale models when interior point methods encounter difficulties.
New Features
- Model Re-optimisation: The Knitro API has been updated to allow re-optimisation of a model after modifications to quadratic and non-linear structures.
- Configurable Cut-off for Branch-and-Bound: New parameters have been introduced to set absolute and relative cut-off levels in the Branch-and-Bound algorithm, offering better control over optimisation processes.
Improved Performance
-
Knitro 15.0 enhances the efficiency and robustness of existing optimisation methods:
- Faster Resolution of MILP and MINLP Problems: The addition of new presolve operations allows for faster resolution of your mixed-integer programming models, particularly for non-linear problems (MINLP) with linear structures.
- Faster Hessian Approximation: This version offers significant performance improvements for Hessian approximation with the limited-memory Broyden-Fletcher-Goldfarb-Shanno method, especially for large-scale problems.
- Use of Apple Accelerate Sparse Solvers Library: Users of Mac-ARM platforms can now use Apple Accelerate’s specific solvers for sparse matrix equation systems, thereby improving performance on large-scale models.
Thanks to these improvements, Knitro 15.0 reduces solution times on continuous problems on our pooling and LP datasets by 46% and 14% respectively, compared with version 14.2. Our evaluation datasets are composed from academic and customer cases instances, with application in water networks, oil and gas, refinery processing, etc.
Knitro 15.0 is also more robust on discrete problems, with new instances solved on our MINLP convex dataset. On the latter, the primal-dual integral gap is reduced by more than 10%, demonstrating increased convergence speed and solution quality.
More Accessible Interfaces
- Python Interface Available on PyPI: The installation process for Knitro’s Python interface has been completely revamped. It is now possible to install Knitro and its Python interface directly from PyPI, the Python package manager, greatly facilitating the use of our non-linear solver from this language.
- Java Interface Compatible with Mac ARM: With this new version, Knitro’s Java interface is now available for users of Mac ARM platforms.
To learn more about Knitro 15.0 and its capabilities, visit the official documentation or review the detailed release notes.
Try Knitro 15.0
Artelys Knitro 15.0 is now available for download on Artelys website from your customer area. Users can explore the software’s features through a free trial or contact the Artelys team for licensing and pricing details.
Use-cases
Discover the use-cases that transform your ideas into practical realities, building a future where technology meets real needs in an exceptional way with Knitro.
Artelys Introduces Future Sight: a Visualisation Tool Supporting the Energy Transition
As partner in the European Climate + Energy Modeling Forum (ECEMF) – a Horizon 2020 Europe project uniting research institutes and leading energy modellers in Europe – Artelys has contributed to modeling activities powered by Artelys Crystal Super Grid modelisation tool, and has led the development of a fully-fledged visualisation tool.
Supercharging Optimisation: How Artelys, FICO and NVIDIA cuOpt Join Efforts to Scale Up Energy System Optimisation
As energy system models continue to scale—reaching up to hundreds of millions of variables and constraints—traditional CPU-based optimisation solvers are hitting performance and memory bottlenecks. These increasingly complex models are essential for planning Europe’s energy transition, yet solving them within realistic timeframes has become a pressing challenge.
Artelys is powering SEleNe CC’s Common Grid Model (CGM) Service
Artelys provides maintenance and support services of the calculation module used to perform the European Merging Function (EMF) at SEleNe CC
Improving electrical transformer production planning with artificial intelligence solutions
Improving high-voltage transformer production planning at the Hitachi Energy factory in Varennes with advanced artificial intelligence solutions.
subscribe to our newsletters
© ARTELYS document.write(new Date().getFullYear()); • All rights reserved • Legal mentions