The Role of Measurement Uncertainty in Numeric Law Induction

Empirical induction of numeric laws aims at nding numerical laws from an experimentation set, using symbolic and inductive techniques. Faced with experimental data, these techniques must be adapted in order to deal with noisy data. In this article, we present the solutions for dealing with numerical uncertainty which are implemented in our system ARC 16]. These solutions are inspired by experimental science and join simplicity to eeciency. Moreover, when used in a more realistic way, numerical uncertainty can signiicantly improve not only the accuracy of the results but also the search algorithm's eeciency.