Ade Fajemisin and Donato Maragno on recent trends and challenges in applying ML to optimization problems.
This week’s guests are Ade Fajemisin (Postdoctoral Researcher) and Donato Maragno (PhD Student) of the University of Amsterdam. They were co-authors of a recent paper (“Optimization with Constraint Learning: A Framework and Survey”) that explores how machine learning can be used to learn constraints in optimization problems. Our conversation focused on key findings in their paper, as well as trends in the use of machine learning for optimization problems.
Optimization problems are routine in many industries including logistics, manufacturing, retail, energy, and financial services. Such problems can often be complex and hard to solve in practice, and as such teams often have to solve simpler and perhaps less realistic optimization tasks. In previous episodes, I’ve had guests describe how reinforcement learning can be used to tackle large-scale simulation and optimization problems.
Highlights in the video version:
- Introduction to Ade Fajemisin and Donato Maragno
Why focus on machine learning for optimization?
Constraint learning and machine learning for organization
State of tools for optimization
What ML techniques are useful for optimization?
Reinforcement learning for optimization and simulation problems
Practitioners working on ML optimization
The optimization community
Tools, frameworks, and packages
- A video version of this conversation is available on our YouTube channel.
- Nic Hohn and Max Pumperla: Reinforcement Learning in Real-World Applications
- Jack Clark: The 2022 AI Index
- Barret Zoph and Liam Fedus: Efficient Scaling of Language Models
- What Is Graph Intelligence?
- Applications of Reinforcement Learning: Recent examples from large US companies
[Image: Dell Rapids grain elevator by Ben Lorica.]