Kasra Jalaldoust


kasra@cs.columbia.edu


CS@Columbia



Partial Transportability for Domain Generalization


NeurIPS2024


Kasra Jalaldoust, Alexis Bellot, Elias Bareinboim
Proceedings of the 38th Annual Conference on Neural Information Processing Systems., 2024 May

technical report
Cite

Cite

APA   Click to copy
Jalaldoust, K., Bellot, A., & Bareinboim, E. (2024). Partial Transportability for Domain Generalization. Proceedings of the 38th Annual Conference on Neural Information Processing Systems.


Chicago/Turabian   Click to copy
Jalaldoust, Kasra, Alexis Bellot, and Elias Bareinboim. Partial Transportability for Domain Generalization. Proceedings of the 38th Annual Conference on Neural Information Processing Systems., May 2024.


MLA   Click to copy
Jalaldoust, Kasra, et al. Partial Transportability for Domain Generalization. Proceedings of the 38th Annual Conference on Neural Information Processing Systems., May 2024.


BibTeX   Click to copy

@techreport{kasra2024a,
  title = {Partial Transportability for Domain Generalization},
  year = {2024},
  month = may,
  institution = {Proceedings of the 38th Annual Conference on Neural Information Processing Systems.},
  author = {Jalaldoust, Kasra and Bellot, Alexis and Bareinboim, Elias},
  month_numeric = {5}
}

Abstract

A fundamental task in AI is providing performance guarantees for predictions made in unseen domains. In practice, there can be substantial uncertainty about the distribution of new data, and corresponding variability in the performance of existing predictors. Building on the theory of partial identification and transportability, this paper introduces new results for bounding the value of a functional of the target distribution, such as the generalization error of a classifiers, given data from source domains and assumptions about the data generating mechanisms, encoded in causal diagrams. Our contribution is to provide the first general estimation technique for transportability problems, adapting existing parameterization schemes such Neural Causal Models to encode the structural constraints necessary for cross-population inference. We demonstrate the expressiveness and consistency of this procedure and further propose a gradient-based optimization scheme for making scalable inferences in practice. Our results are corroborated with experiments.