Interactive optimization prototype
ANTS
Compare algorithms on the same assignment problem
The challenge
When there are many agents and tasks, the assignment depends on the method used. The aim was to make that difference visible: use one cost matrix and compare the assignments and outcomes produced by different algorithms.
Our contribution
Andrey built a Flask application with matrices and charts comparing the Hungarian algorithm, ant colony optimisation, simulated annealing and Sinkhorn approximation, using NumPy and SciPy.
The outcome
The interactive application turns an abstract optimisation problem into a readable experiment. The same data can be used to compare four methods, inspect matrices and charts and discuss why the algorithms behave differently.