Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
41 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
41 tokens/sec
o3 Pro
7 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Reduced-Rank Multi-objective Policy Learning and Optimization (2404.18490v1)

Published 29 Apr 2024 in cs.LG and stat.ML

Abstract: Evaluating the causal impacts of possible interventions is crucial for informing decision-making, especially towards improving access to opportunity. However, if causal effects are heterogeneous and predictable from covariates, personalized treatment decisions can improve individual outcomes and contribute to both efficiency and equity. In practice, however, causal researchers do not have a single outcome in mind a priori and often collect multiple outcomes of interest that are noisy estimates of the true target of interest. For example, in government-assisted social benefit programs, policymakers collect many outcomes to understand the multidimensional nature of poverty. The ultimate goal is to learn an optimal treatment policy that in some sense maximizes multiple outcomes simultaneously. To address such issues, we present a data-driven dimensionality-reduction methodology for multiple outcomes in the context of optimal policy learning with multiple objectives. We learn a low-dimensional representation of the true outcome from the observed outcomes using reduced rank regression. We develop a suite of estimates that use the model to denoise observed outcomes, including commonly-used index weightings. These methods improve estimation error in policy evaluation and optimization, including on a case study of real-world cash transfer and social intervention data. Reducing the variance of noisy social outcomes can improve the performance of algorithmic allocations.

PDF HTML Abstract
Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Bookmark Chat Bubble Oval Streamline Icon: https://streamlinehq.com Chat (Pro)
Definition Search Book Streamline Icon: https://streamlinehq.com
References (33)
  1. S. Anand and A. Sen. Concepts or human development and poverty: a multidimensional perspective. United Nations Development Programme, Poverty and human development: Human development, pages 1–20, 1997.
  2. Policy learning with observational data. Econometrica, 2020.
  3. The surrogate index: Combining short-term proxies to estimate long-term treatment effects more rapidly and precisely. Working Paper 26463, National Bureau of Economic Research, November 2019. URL http://www.nber.org/papers/w26463.
  4. Unpacking a multi-faceted program to build sustainable income for the very poor. Journal of Development Economics, 155:102781, 2022. ISSN 0304-3878. doi: https://doi.org/10.1016/j.jdeveco.2021.102781. URL https://www.sciencedirect.com/science/article/pii/S0304387821001401.
  5. Doubly robust estimation in missing data and causal inference models. Biometrics, 2005.
  6. Hamsa Bastani. Predicting with proxies: Transfer learning in high dimension. Manage. Sci., 67(5):2964–2984, may 2021. ISSN 0025-1909. doi: 10.1287/mnsc.2020.3729. URL https://doi.org/10.1287/mnsc.2020.3729.
  7. (machine) learning what policies value. arXiv preprint arXiv:2206.00727, 2022.
  8. The returns to microenterprise support among the ultra-poor: A field experiment in post-war uganda. Working Paper 21310, National Bureau of Economic Research, June 2015. URL http://www.nber.org/papers/w21310.
  9. Treatment policy learning in multiobjective setting with fully observed outcomes. In Proceedings of the 26th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD ’20), 2020.
  10. Tackling psychosocial and capital constraints to alleviate poverty. Nature, 605(7909):291–297, 2022a.
  11. Tackling psychosocial and capital constraints to alleviate poverty. Nature, 605(7909):291––97, 2022b. doi: 10.1038/s41586-022-04647-8.
  12. Optimal selection of reduced rank estimators of high-dimensional matrices. 2011.
  13. Learning individualized treatment rules for multiple-domain latent outcomes. Journal of the American Statistical Association, 116, 2021.
  14. Practical policy optimization with personalized experimentation, 2023.
  15. Some new perspectives on the method of control variates. Department of Management Science and Engineering, Stanford University, Stanford CA 94305, USA, 2002.
  16. Estimating scaled treatment effects with multiple outcomes. Statistical methods in medical research, 28(4):1094–1104, 2019.
  17. Testing causal theories with learned proxies. Annual Review of Political Science, 25(1):419–441, 2022. doi: 10.1146/annurev-polisci-051120-111443. URL https://doi.org/10.1146/annurev-polisci-051120-111443.
  18. Winston Lin. Agnostic notes on regression adjustments to experimental data: Reexamining freedman’s critique. Annals of Applied Statistics, 7:295–318, 2013. doi: https://doi.org/10.1214/12-AOAS583.
  19. Linear fitted-q iteration with multiple reward functions. The Journal of Machine Learning Research, 13:3253–3295, 2012.
  20. Estimation and optimization of composite outcomes. The Journal of Machine Learning Research, 22:1–40, 2021.
  21. Machine-learning tests for effects on multiple outcomes. arXiv preprint arXiv:1707.01473, 2017.
  22. David J. McKenzie. Measuring inequality with asset indicators. Journal of Population Economics, 18:229–260, 2005. doi: https://doi.org/10.1007/s00148-005-0224-7.
  23. Multivariate reduced rank regression: Theory and applications. Springer, 1998.
  24. Multivariate reduced-rank regression: theory, methods and applications, volume 225. Springer Nature, 2022.
  25. Off-policy evaluation for large action spaces via embeddings. In Kamalika Chaudhuri, Stefanie Jegelka, Le Song, Csaba Szepesvari, Gang Niu, and Sivan Sabato, editors, Proceedings of the 39th International Conference on Machine Learning, volume 162 of Proceedings of Machine Learning Research, pages 19089–19122. PMLR, 17–23 Jul 2022. URL https://proceedings.mlr.press/v162/saito22a.html.
  26. Measuring the predictability of life outcomes with a scientific mass collaboration. Proceedings of the National Academy of Sciences, 117(15):8398–8403, 2020.
  27. Adjusting for nonignorable drop-out using semiparametric nonresponse models. Journal of the American Statistical Association, 94(448):1096–1120, 1999. doi: 10.1080/01621459.1999.10473862. URL https://www.tandfonline.com/doi/abs/10.1080/01621459.1999.10473862.
  28. Probabilistic principle component analysis. Journal of Royal Statistical Society, 61, 1999.
  29. Choosing a proxy metric from past experiments. arXiv preprint arXiv:2309.07893, 2023.
  30. (when) should you adjust inferences for multiple hypothesis testing? arXiv preprint arXiv:2104.13367, 2021.
  31. Jeffrey M. Wooldridge. Econometric analysis of cross section and panel data. MIT Press, 2001.
  32. Efficient heterogeneous treatment effect estimation with multiple experiments and multiple outcomes. arXiv preprint arXiv:2206.04907, 2022.
  33. Dimension reduction and coefficient estimation in multivariate linear regression. Journal of the Royal Statistical Society. Series B (Statistical Methodology), 69(3):329–346, 2007. ISSN 13697412, 14679868. URL http://www.jstor.org/stable/4623272.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (3)
  1. Ezinne Nwankwo (3 papers)
  2. Michael I. Jordan (438 papers)
  3. Angela Zhou (23 papers)
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets