All models are wrong, but some
Machine-Learned
models are useful
Enrico Camporeale
, University of Colorado at Boulder, Boulder, United States