Responsive image

Flavio du Pin Calmon

Assistant Professor

I am an Assistant Professor of Electrical Engineering at Harvard's John A. Paulson School of Engineering and Applied Sciences. Before joining Harvard I was a social good post-doctoral fellow at IBM Research in Yorktown Heights, New York. I received my Ph.D. in Electrical Engineering and Computer Science at MIT. My main research interests are information theory, signal processing, and machine learning.

Research

My research group works on information theory, signal processing, and machine learning. We aim to develop foundational theory that guides the design of machine learning algorithms that do not cause harm. Recently, our work has centered around three critical challenges in machine learning: fairness, privacy, and reliability. Our vision is that information theory is key for the responsible design of machine learning systems and that a theory-guided approach can massively outperform heuristics. You can find more details about our research in the selected publications below.

I consider myself a scientist who is an engineer at heart, and I enjoy doing fundamental research that serves as a design driver for practical applications. My research adopts the information-theoretic blueprint: for a given performance metric (e.g., precision, fairness, privacy), I ask not "How can a given system be made better?" but instead "What is the absolute best a system can be?" I use information theory to characterize fundamental performance limits of machine learning systems, and then design algorithms that try to achieve these limits.

As a Brazilian-American engineer, I am passionate about broadening the participation of students from diverse backgrounds, including from Latin-America, in information theory and machine learning. I am also excited about addressing data-driven challenges specific to the developing world. If this is something that interests you as well, please get in touch!

My research group is supported by the NSF and by generous gifts and awards from Google, IBM, Amazon, and Oracle Research.

Recent announcements

We are hiring! Estamos contratando!

There is a post-doc position and Ph.D. positions avaiable in my group for the upcoming year. Please reach out for more information.

  • September, 2022 — Symposium on Machine Learning and Information Theory at CNMAC 2022

    I organized a symposium at the 2022 Congresso Nacional de Matemática Aplicada e Computacional in Campinas, São Paulo, Brazil. The workshop was supported in part by a Harvard Brazil Lemann Research Award.

  • August, 2022 — Hsiang Hsu's research on predictive multiplicity and the Rashomon effect featured as the Meta Research PhD Fellowship Spotlight!
  • July, 2022 — Tutorial at ISIT 2022 on Information-Theoretic Tools for Responsible Machine Learning

    I organized a tutorial at ISIT 2022 on responsible ML together with Haewon Jeong, Shahab Asoodeh, and Mario Diaz. Slides can be found here!

  • May, 2022 — Congratulations Dr. Hao Wang on defending your Ph.D.!

    And congratulations for joining the MIT-IBM research lab in Cambridge! You can read Hao's Thesis here.

  • July, 2021 — Haewon's research featured at SEAS!

    Read more about Haewon Jeong's exciting work at the SEAS website.

  • June, 2021 — NSF-Amazon Fairness in AI Award.

    We just kicked-off the research sponsored by a recent NSF and Amazon FAI grant! Very grateful to the National Science Foundation and Amazon for supporting our work.

  • May, 2021 — Congratulations Hsiang Hsu for becoming a Meta Ph.D. Fellow!

    The annoucnement can be found here.

  • April, 2021 — Título de Honra ao Mérito da Universidade de Brasília.

    Extremely honored for receiving the inaugural Título de Honra ao Mérito (Honor to the Merit award) from the University of Brasilia (UnB), my alma mater. This new award recognizes alumni who have achieved national and/or international recognition in their field of study at UnB. I was the first alumni chosen for the award from the area of "technology and exact sciences" (i.e., engineering, math, statistics, and CS).

  • July, 2020 — Teaching commendation.

    Honored for receiving a commendation from the Harvard College's Dean of Undergraduate Education for "extraordinary teaching during extraordinary times." Signals & Communications (ES 156) is more important than ever!

  • June, 2020 — Amazon Research Award.

    Alexa, thank you for sponsoring our work with an Amazon Research Award!

  • September, 2019 — NSF Grants.

    I am once again grateful to National Science Foundation for supporting our work on privacy with an NSF Medium grant and in fair machine learning with an NSF Eager grant.

  • March, 2019 — Google Faculty Research Award.

    Ok Google, thank you so very much for your generous gift (see under "Machine Learning and Data Mining").

  • Dec, 2018 — NSF CAREER Award!

    I am very grateful for the support of the National Science Foundation for our research on information-theoretic foundations of fair machine learning (check out this GSAS feature). You can learn more details about the award at the Harvard SEAS website.

  • Nov, 2018 —IBM Open Collaborative Research Award.

    Thank you IBM!

  • April, 2018 —Lemann Brazil Research Fund Award.

    We are excited to organize a course on ML at FEEC/Unicamp in August 2019! You can find more details here.

    • Research Group

    I am very fortunate to work with an amazing group of students, post-docs, and visitors.

    Current Members

  • Wael Alghmadi (PhD, G6)
  • Hsiang Hsu (PhD, G6)
  • Juan Felipe Gomez (PhD, G3)
  • Carol Long (PhD, G2)
  • Lucas Monteiro Paes (PhD, G2)
  • Alex Oesterling (PhD, G1, co-advised with Hima Lakkaraju)
  • Bogdan Kulynych (visiting graduate student from EPFL)

    • Alumni

  • Hao Wang: Former graduate student (Ph.D.), now at the MIT-IBM Watson AI Lab
  • Haewon Jeong: Former Post-Doc, now faculty at UCSD
  • Shahab Asoodeh: Former Post-Doc, now faculty at McMaster University
  • Juwendo Denis: Former Post-Doc, next stop Senior Algoirthm Designer at Parallel Wireless
  • Madeleine Barowsky: Former graduate student (M.Sc.), now software engineer at Etsy
  • Berk Ustun: Former CRCS Post-Doc, now faculty at UCSD
  • Mario Diaz: Former Post-Doc (joint with ASU), now faculty at UNAM
  • Javier Zazo: Former CRCS Post-Doc, next stop Microsoft Research

    • Selected Publications

    An (almost) up-to-date list of publications can be found on Google Scholar.

    The list below includes a few selected recent publications and pre-prints.
  • Generalization bounds for noisy iterative algorithms using properties of additive noise channels
    H. Wang, R. Gao, and F. P. Calmon
    Journal of Machine Learning Research (Accepted)
  • Beyond Adult and COMPAS: Fair Multi-Class Prediction via Information Projection
    W. Alghamdi, H. Hsu, H. Jeong, H. Wang, P.W. Michalak, S. Asoodeh, and F. P. Calmon
    Accepted to the Thirty-sixth Conference on Neural Information Processing Systems (NeurIPS), 2022 (Oral Presentation)
  • Rashomon Capacity: A Metric for Predictive Multiplicity in Probabilistic Classification
    H. Hsu and F. P. Calmon
    Accepted to the Thirty-sixth Conference on Neural Information Processing Systems (NeurIPS), 2022
  • On the Epistemic Limits of Personalized Prediction
    L.M. Paes, C.X. Long, B. Ustun, and F. P. Calmon
    Accepted to the Thirty-sixth Conference on Neural Information Processing Systems (NeurIPS), 2022
  • Measuring Information from Moments
    W. Alghamdi and F.P. Calmon
    IEEE Trans. Inf. Theory, 2022 (to appear).
  • Fairness without imputation: A decision tree approach for fair prediction with missing values
    H. Jeong, H. Wang, and F.P. Calmon
    Proceedings of the AAAI Conference on Artificial Intelligence, vol. 36, issue 9, pp. 9558-9566, 2022 (Oral Presentation).
  • Cactus mechanisms: Optimal differential privacy mechanisms in the large-composition regime
    W. Alghamdi, S. Asoodeh, F. P. Calmon, O. Kosut, L. Sankar, and F. Wei
    2022 IEEE International Symposium on Information Theory (ISIT).
  • The Saddle-Point Accountant for Differential Privacy
    W. Alghamdi, S. Asoodeh, F. P. Calmon, J.F. Gomez, O. Kosut, L. Sankar, and F. Wei
    arXiv pre-print (under review).
  • Analyzing the generalization capability of SGLD using properties of gaussian channels
    H. Wang, Y. Huang, R. Gao, and F.P. Calmon
    Advances in Neural Information Processing Systems 34 (NeurIPS 2021), pp. 24222-24234, 2021
  • Generalizing Correspondence Analysis for Applications in Machine Learning
    H. Hsu, S. Salamatian, and F.P. Calmon
    IEEE Trans. on Pattern Matching and Machine Intelligence, 2022 (to appear).
  • Optimized Score Transformation for Consistent Fair Classification
    D. Wei, K. N. Ramamurthy, and F. P. Calmon
    Journal of Machine Learning Research, vol. 22(258), pp. 1 - 78, 2021.
  • To Split or Not to Split: The Impact of Disparate Treatment in Classification
    H. Wang, H. Hsu, M. Diaz, and F. P. Calmon
    IEEE Trans. Inf. Theory, vol. 67, no. 10, pp. 6733 - 6757, April 2021.
  • ε-Approximate Coded Matrix Multiplication Is Nearly Twice as Efficient as Exact Multiplication
    H. Jeong, A. Devulapalli, V.R. Cadambe, F.P. Calmon
    IEEE Journal on Selected Areas in Information Theory, vol. 2, no. 3, pp. 845-854, 2021
  • Predictive Multiplicity in Classification
    C. T. Marx, F. P. Calmon, and B. Ustun
    Int. Conf. on Machine Learning (ICML), 2020

    • Patents

    Teaching

    Fall 2022: Information Theory (ES 250)

    Spring 2022: Covid teaching relief

    Fall 2021: Information Theory (ES 250)

    Spring 2021: Signals & Communications (ES 156)

    Fall 2020: On parental leave from teaching

    Spring 2020: Signals & Communications (ES 156)

    Fall 2019: Information Theory (ES 250)

    Spring 2019: Signals and Systems (ES 156)

    Spring 2018: Signals and Systems (ES 156)

    Fall 2017: Information Theory (ES 250)

    Contact

    Email: flavio (at) seas (dot) harvard (dot) edu

    Office:
    Science & Engineering Complex
    150 Western Ave
    Office 3.306
    Allston, MA 02134