• Corpus ID: 225709808

A better method to enforce monotonic constraints in regression and classification trees

@article{Auguste2020ABM,
  title={A better method to enforce monotonic constraints in regression and classification trees},
  author={Charles Auguste and Sean Malory and Ivan Smirnov},
  journal={ArXiv},
  year={2020},
  volume={abs/2011.00986}
}
In this report we present two new ways of enforcing monotone constraints in regression and classification trees. One yields better results than the current LightGBM, and has a similar computation time. The other one yields even better results, but is much slower than the current LightGBM. We also propose a heuristic that takes into account that greedily splitting a tree by choosing a monotone split with respect to its immediate gain is far from optimal. Then, we compare the results with the… 
Robust and Provably Monotonic Networks
TLDR
A new method to constrain the Lipschitz constant of dense deep learning models that can also be generalized to other architectures is presented that is minimally constraining and allows the underlying architecture to maintain higher expressiveness compared to other techniques.

References

Figure 12: Penalty function for different penalization parameters