Timezone: »

 
Poster
Multidimensional Shape Constraints
Maya Gupta · Erez Louidor · Oleksandr Mangylov · Nobu Morioka · Taman Narayan · Sen Zhao

Wed Jul 15 11:00 AM -- 11:45 AM & Wed Jul 15 10:00 PM -- 10:45 PM (PDT) @ None #None
in Poster Session 25 »

We propose new multi-input shape constraints across four intuitive categories: complements, diminishers, dominance, and unimodality constraints. We show these shape constraints can be checked and even enforced when training machine-learned models for linear models, generalized additive models, and the nonlinear function class of multi-layer lattice models. Toy examples and real-world experiments illustrate how the different shape constraints can be used to increase interpretability and better regularize machine-learned models.

Keywords: Supervised Learning · Accountability, Transparency and Interpretability · Other · Safety

Author Information

Maya Gupta (Google)
Erez Louidor (Google)
Oleksandr Mangylov (Google Research)
Nobu Morioka (Google Research)
Taman Narayan (Google)
Sen Zhao (Google Research)

More from the Same Authors