Finding the best decision can be difficult when an optimization problem contains many competing possibilities and a landscape of local solutions. Can machine learning help the algorithms doing that search make better choices?

Bissan Ghaddar explored this question in a Sargent Centre seminar on 24 February 2026. Her talk examined how learning can improve components of nonlinear optimization solvers, from choosing where to branch in a search tree to generating constraints that tighten mathematical approximations.

The seminar presented benchmark experiments comparing learning-based branching and constraint generation with standard approaches. The work connects machine learning with optimization: learning guides parts of the search, while mathematical models continue to define the decision problem.

Official event details and programme.

Front photograph: Shadowssettle, Imperial College Central Library and Queen’s Tower, CC BY-SA 4.0. Cropped within the Datalab card.

Bissan Ghaddar is a member of IE Research Datalab at IE University.