Discrete Optimization Challenges in Computational Sustainability
- 14:00 10th October 2018 ( week 1, Michaelmas Term 2018 )Lecture Theatre B, Wolfson Building
Computational sustainability is a new interdisciplinary research focused on computational problems that arise in the quest for sustainable development. In this talk, I will provide a sample of computational sustainability problems from the areas of biodiversity conservation and infrastructure planning. I will describe, for example, network design problems motivated by challenging planning problems in wildlife corridor conservation, as well as our system PAWS-ML for informing ranger patrol planning for anti-poaching efforts.
I will emphasize the unique opportunities for scalable constraint reasoning and optimization techniques to contribute to the new research area of computational sustainability. I will describe our recent advances in improving the state-of-the-art in large-scale optimization by leveraging machine learning techniques to inform the design of combinatorial search algorithms. These include ML-driven advances in Branch-and-Bound search for Mixed Integer Programming and in greedy algorithms for Graph optimization problems.