Latest release enables data professionals to easily integrate machine learning models into optimization models to solve new types of problems.
Gurobi Optimization, LLC, the leader in decision intelligence technology, announced the release of Gurobi Optimizer 10.0. This release provides customers with a boost to its already industry-leading speed, the ability to embed machine learning models directly into Gurobi optimization models, and new tools for model development, monitoring, and advanced diagnosis—so users can solve new types of problems, even faster than before.
AI News: Infobip Creates AI-powered Chatbot for Uber
“We have the absolute best minds in optimization here at Gurobi. Across every department, you’ll find people who aren’t just smart—they’re also deeply committed to our customers and to providing the best possible experience. I’m proud to be a part of this team.”
Performance Improvements and Advanced Solving Techniques
The Gurobi R&D team continues to push the boundaries of performance—resulting in improvements to existing algorithms and the development of several brand-new techniques. As a result, Gurobi Optimizer 10.0 has achieved the following performance improvements since the release of Gurobi Optimizer 9.5:
“We’ve achieved a more than 75x speedup on MILP since version 1.1. But more importantly, Gurobi 10.0 can now solve even more models easily, including some models that were, until now, intractable,” explained Dr. Tobias Achterberg, Vice President of Research and Development at Gurobi Optimization.
Gurobi 10.0 also includes the following advances in the underlying algorithmic framework:
- New network simplex algorithm – Greatly speeds up solving LPs with network structure.
- New heuristic for QUBO models, which can arise in quantum optimization – Improves Gurobi’s ability to quickly find good feasible solutions for quadratic unconstrained Boolean optimization problems.
- Significant performance gains on MIPs that contain machine learning models – Results in a more than 10x improvement on certain models that contain embedded neural networks with ReLU activation functions.
- New optimization-based bound tightening (OBBT) algorithm – Greatly speeds up solving nonconvex MIQCP models.
- Reorganized concurrent LP solver – Improves performance and reduces memory footprint.
Innovative Data Science Integration
With Gurobi Machine Learning—an open-source Python project to embed trained machine learning models directly into Gurobi—data scientists can more easily tap into the power of mathematical optimization.
Latest Aithority Insights : NVIDIA Raises the Standard of Low Code DevOps with the NVIDIA AI Enterprise 2.1
Specifically, Gurobi Machine Learning allows users to add a trained machine learning model as a constraint to a Gurobi model (e.g., from scikit-learn, TensorFlow/Keras, or PyTorch). Thus, users can estimate a real-world system by training a machine learning model, and then use this machine learning model as a constraint in Gurobi, in order to optimize controls on that system.
“We’re aiming to connect the world of data science with the world of optimization. With Gurobi, you can take your machine learning ‘black box’ that’s generating your predictions and plug it directly into your optimization model—enabling you to connect your forecasting with optimization,” explained Achterberg.
With this release, we’re also making it more convenient to integrate gurobipy model building with pandas objects through a new, dedicated open-source package. (Available on GitHub/PyPI in Q4 2022.)
Enterprise Development and Deployment Experience
To make its solver even more accessible and easy to use, the Gurobi team has integrated new tools for model development, monitoring, and advanced diagnosis:
- Significant enhancements to the matrix-friendly API in gurobipy – All matrix-friendly modeling objects now support multiple dimensions, and dimension handling leans consistently on NumPy, including broadcasting.
- New logistic general constraint – Makes it easy to incorporate a constraint in MIP that models the logistic function.
- NuGet package for .NET – Allows .NET users to download Gurobi directly from the NuGet server.
- Memory limit parameter that allows graceful exit – Users can set a memory limit and still get the best solution and resume the optimization after the limit was hit.
- New Compute Server dashboards – The Gurobi Compute Server now includes two new dashboards, enabling users to monitor metrics over time and drill down to the actual activity to better understand the cluster usage and application behavior.
- Expanded platform support – Gurobi 10.0 includes support for Python 3.11 and Linux on ARM 64-bit.
Gurobi introduced its Web License Service (WLS) for Docker and Kubernetes container environments last year, with the release of Gurobi 9.5. With Gurobi 10.0, the team has expanded WLS to support nearly all types of containerized environments. Moreover, customers can now also obtain WLS licenses that allow them to run Gurobi in virtually all deployment scenarios, including containerized environments, virtual machines, and bare-metal machines, across Linux, macOS, and Windows.
“Our customers love our WLS and the flexibility it provides. And now they can dynamically deploy Gurobi software in even more environments,” explained Duke Perrucci, Gurobi’s Chief Operating Officer.
Additionally, starting with Gurobi 10.0, major product releases—and their subsequent minor and technical product releases—will be supported for a term of three years from the initial major product release date. For example, Gurobi version 10.0.0 (released in November 2022) and minor releases between 10.0 and 11.0 will be supported until November 2025.
“This helps create predictability for our customers, so they know exactly how long a version will be supported,” explained Dr. Sonja Mars, Director of Optimization Support at Gurobi Optimization. “We aim to deliver expert technical guidance and support for our customers—and this policy helps eliminate the guesswork. We want our customers to get the help they need, when they need it.”
Dr. Edward Rothberg, Chief Executive Officer and Co-founder of Gurobi Optimization added, “We have the absolute best minds in optimization here at Gurobi. Across every department, you’ll find people who aren’t just smart—they’re also deeply committed to our customers and to providing the best possible experience. I’m proud to be a part of this team.”
Latest NaturalAI Insights: InspireXT Announces Acquisition Of NaturalAI – A Conversational Artificial Intelligence Platform To Expand Its Solution Portfolio
[To share your insights with us, please write to sghosh@martechseries.com]
Comments are closed.