February 12, 2020

220 words 2 mins read

How machine learning meets optimization

How machine learning meets optimization

Machine learning and constraint-based optimization are both used to solve critical business problems. They come from distinct research communities and have traditionally been treated separately. But Jari Koister examines how they're similar, how they're different, and how they can be used to solve complex problems with amazing results.

Talk Title How machine learning meets optimization
Speakers Jari Koister (FICO )
Conference Strata Data Conference
Conf Tag Make Data Work
Location New York, New York
Date September 24-26, 2019
URL Talk Page
Slides Talk Slides

Optimization and ML are increasingly intersecting, occasionally overlapping, sometimes complementary, and most often best used in combination. Data scientists should be interested in operations research, and operations researchers are increasingly using machine learning. Jari Koister dives into understanding and applying these two techniques. He explores when optimization techniques originating from operations research are the better solution and when it’s beneficial to apply ML. More importantly, he outlines how complex, high-value business problems can be better solved by combining the techniques rather than by using only one of them. People struggle to describe how they relate from a theoretical perspective, if ML is just an optimization problem, or how simplex, MIP, interior point, neural networks, and gradient decent relate. Jari outlines a model that helps you understand how the two techniques are related, overlap, and differ.

comments powered by Disqus