Intel and National Science Foundation Fund Research to Apply AI to Optimize Large-Scale Problems

Highlights

  • Intel and the U.S. National Science Foundation are jointly funding the AI Institute for Advances in Optimization, investing $20 million in a five-year research initiative starting in 2021.

  • Researchers will develop machine learning tools to solve large-scale optimization problems that were previously intractable.

author-image

Von

Intel and the U.S. National Science Foundation (NSF) have partnered to fund the AI Institute for Advances in Optimization, investing $20 million in the five-year research initiative to develop breakthrough optimization methods and applications using machine learning. The institute will bring together researchers in computer science, operations research, machine learning (ML), and related fields to develop new AI tools for solving previously “impossible” large-scale problems in domains such as resource allocation, planning, system design and optimization, and hardware and software design.

Proposal submissions are due by December 4, 2020 at the AI Institute for Advances in Optimization. The award winners and project descriptions will be announced in Q3 2021.

Emerging optimization methods rely on data-driven approaches found in machine learning, which could enable a broad array of applications, such as in automated machine learning, planning and transportation, and system-level optimizations.

“By harnessing machine learning with optimization techniques, the AI Institute for Advances in Optimization will enable AI to take on mission-critical problems of unprecedented scale and complexity,” said Henry Kautz, division director of Information & Intelligent Systems (CISE/IIS) at NSF.

However, breakthroughs in optimization will require foundational advancements and new perspectives. For example, optimizing chip design require algorithms to operate in large search spaces. Tackling resource planning requires reasoning in previously unseen environments. Researchers will explore ways to integrate perspectives from both classical constrained and unconstrained optimization, and modern data-driven approaches such as reinforcement learning. These methods, along with recent theoretical breakthroughs in AI, computer science, and operations research, will enable highly efficient combinatorial search to tackle foundational challenges.

“Intel is committed to collaborating with the NSF to support the university research community in moving AI forward. We’re excited to see the ingenuity and innovation that researchers will apply to large-scale problems,” said Gabriela Cruz Thompson, director of university research and collaborations at Intel Labs.