Globally Optimal Symbolic Regression

Vernon Austel , Sanjeeb Dash (1) , Oktay Gunluk , Lior Horesh , Leo Liberti (2) , Giacomo Nannicini (2) , Baruch Schieber (3)

1 IBM Watson Research Center
2 LIX - Laboratoire d'informatique de l'École polytechnique [Palaiseau]
3 IBM Thomas J. Watson Research Center

In this study we introduce a new technique for symbolic regression that guarantees global optimality. This is achieved by formulating a mixed integer non-linear program (MINLP) whose solution is a symbolic mathematical expression of minimum complexity that explains the observations. We demonstrate our approach by rediscovering Kepler's law on planetary motion using exoplanet data and Galileo's pendulum periodicity equation using experimental data.

Domaines: Recherche opérationnelle [math.OC]; Intelligence artificielle [cs.AI]; Apprentissage [cs.LG]

Vernon Austel, Sanjeeb Dash, Oktay Gunluk, Lior Horesh, Leo Liberti, et al.. Globally Optimal Symbolic Regression. Interpretable ML, satellite workshop of NIPS 2017, 2017, Long Beach, United States. ⟨hal-02105286⟩