Postdoc at ASU/Harvard on ML and Privacy

A joint postdoc position at Arizona State University and Harvard University in the area of machine learning and privacy is available immediately. The successful candidate will be working with the research groups of Prof. Lalitha Sankar and Prof. Flavio du Pin Calmon.

Specific topics of focus are the interplay of machine learning and privacy with focus on both rigorous information-theoretical results as well as practical design aspects.

The appointment will be for a period of 12 months initially, with a possibility for renewal. We are looking for strong applicants with an excellent theoretical background and a proven capacity for high-quality research in form of a strong publication record. Knowledge of privacy literature is desirable.

Interested applicants should submit a current CV, a 1-page research statement and a list of three references. Candidates should contact us via email at lsankar@asu.edu and/or flavio@seas.harvard.edu.

DIMACS Workshop on Distributed Optimization, Information Processing, and Learning

My colleague Waheed Bajwa, Alejandro Ribeiro, and Alekh Agarwal are organizing a Workshop on Distributed Optimization, Information Processing, and Learning from August 21 to August 23, 2017 at Rutgers DIMACS. The purpose of this workshop is to bring together researchers from the fields of machine learning, signal processing, and optimization for cross-pollination of ideas related to the problems of distributed optimization, information processing, and learning. All in all, we are expecting to have 20 to 26 invited talks from leading researchers working in these areas as well as around 20 contributed posters in the workshop.

Registration is open from now until August 14 — hope to see some of you there!

Register soon for the 2017 North American School of Information Theory

The site for the 2017 North American School of Information Theory is now live and registration will begin next week. The IT Schools have been going strong for the last few years and are a great resource for students, especially new students, to get some exposure to information theory research beyond their own work and what they learned in class. Most schools do not have several people working on information theory. For students at such institutions the school provides a great way to meet other new researchers in the field.

Why I will no longer do research sponsored by the Department of Homeland Security

Shortly after the inauguration, the new administration announced a blanket ban on  refugees and any entry into the US from 7 proscribed countries. The announcement was met with outrage, ridicule, and legal opposition. Despite its obvious unconstitutionality and complete lack of guidance on implementation, the Department of Homeland Security (DHS) engaged in a weekend of harassment and intimidation at our border. They were big cheerleaders of the effort but also incompetent and unprofessional. At the same time, a pilot project I had worked on, funded by DHS, was about to be submitted for a “phase 2” extension. In light of those events, I asked to be taken off the project entirely and will no longer seek funding from DHS.

Now the administration has revealed an harsh and chilling new plan to terrorize immigrants and their families through a campaign of demonization and deportation. The biggest cheerleader of this plan is the new DHS secretary, John F. Kelly. This only validates my decision: I cannot conduct research under the aegis of an agency whose planned actions run counter to nearly everything I believe. It would be unethical for me to continue to do so. If that were not enough, DHS has also ceded its technical authority on security/privacy issues. I have little confidence that real security/privacy is a priority.

Supporting graduate students in engineering often requires getting funds from multiple sources; it is difficult to sustain a research group (and in some places, get tenure) on NSF (or NIH, in my field at least) funding alone. The usual suspects are industry, foundations, the various defense research programs (ARO, ONR, AFOSR, DARPA), or other government departments (DOE comes to mind). If you work in security/privacy, DHS has also been a (relatively) new source of research funding. Many of the other government agencies do fund basic research but some are focused more on applied research or transitioning research to practice.

There are many considerations to balance. As a theory and algorithms person, the overhead of making “deliverables” can take time away from basic research. Some programs have a lot of check-ins and progress updates which don’t necessarily work with the academic calendar. Finally, trying to describe your theory question about graph topology and its impact on convergence rates of some estimator as crucial to improving “warfighter capabilities” can sometimes feel awkward. My approach thus far has been to try things out. If the proposal is on research I would want to do anyway and everything is public domain and open, then OK. If the experience is not so good/fun, maybe I will consider alternatives. Basic research has lots of applications; improving a control algorithm could yield better drone targeting or more accurate laser-aided surgery.

It was with this in mind that I proceeded with a DHS-funded project on privacy-sensitive anomaly detection. It was a short project but raised a number of interesting new problems and models that I would like to explore, if possible, in the future. It also funded a student whose studies are now threatened by this very administration. Indeed, the entire academic research enterprise in the US is under threat. As academics we can sign petitions, but it is not enough. Scott Aaronson wrote very eloquently about this issue after the initial ban was announced (see also Terry Tao). My department has seen a dramatic decrease in the number of applicants in general and not just from Iran. We were just informed that we can no longer make TA offers for students who are unlikely to get a visa to come here.

The Department of Homeland Security has demonstrated its blatant disregard for moral norms. Why should we trust its scientific norms? What confidence do we have that funding will not be used in some coercive way? What does it say to our students when we ask them to work for DHS? Yes, the government is big, but at some point the argument that it’s mostly the guy at the top who is bad but the rest of the agency is still committed to good science becomes just too hard to swallow. I decided that I can’t square that circle. Each one of us should think hard about whether we want to.

Put José M.F. Moura on the IEEE President Ballot!

José Moura was nominated to be considered as a candidate for IEEE President. The IEEE Board of Directors (BoD) is the body that decides who the candidates will be. They were also behind the proposed IEEE constitutional amendment that was opposed by more than 25 IEEE Societies because it was a rather naked power grab that would have moved much of the constitution to bylaws that could be amended by a small group of individuals, reducing regional representation, and reducing technical activities representation. You can think of it as an attempt by the BoD to Make IEEE Great Again. Prof. Moura spoke up against the amendment (including at an ITSOC BoG meeting) and in the end the amendment did not pass.

It is likely that the BoD’s decision to not nominate him is retaliation for his actions in defense of what he (and many others) felt are the best interests of the IEEE. While this Trump-esque approach to handling dissent may be popular in Washington, it seems quite inappropriate for an international organization such as IEEE. There is a petition at the IEEE website to put him on the ballot. It needs around 4,000 signatures and students members are also welcome to sign. You sign in with your IEEE account and then go to “Annual Election Petitions.” Note this is not a vote for him, but rather allowing him to be on the ballot. Please consider signing!

Tracks: I’m just a soul whose intentions are good

I’m in a bit of a creative rut so I return to the same favorites, many introduced to me by others. This one has a narrative arc though, albeit bittersweet. A present for my cousin.

  1. A Summer Song (Chad & Jeremy)
  2. Golden Age (Chris Staples)
  3. Kaze Wo Atsumete (Happy End)
  4. Soda Shop (Jay Brannan)
  5. Who’s bitten you (Hannah and Sam)
  6. Here Comes a Special Boy (Freezepop)
  7. I Wish You (CAPSULE)
  8. Wedding Daze (Daniel Wilson)
  9. Clay (feat. Marina Price) (Manatee Commune)
  10. I Need A Dollar (Aloe Blacc)
  11. Turn It Around (Lucius)
  12. A Beautiful Woman (Deradoorian)
  13. Temper Temper  (Tuscadero)
  14. Cool Yourself (Thao and the Get Down Stay Down)
  15. (I’ve Got) The Sanctioned Blues (Ultimate Painting)
  16. Get Me Away From Here, I’m Dying (Belle & Sebastian)
  17. Don’t Let Me Be Misunderstood (Nina Simone)
  18. I Like My Face (Miss Erika)
  19. Awkward (Lightning Love)
  20. Ten-Twenty-Ten (Generationals)
  21. Silver Lining (Unplugged) (Stiff Little Fingers)

Postdoctoral Associate at DIMACS

DIMACS, the Center for Discrete Mathematics and Theoretical Computer Science, invites applications for various postdoctoral associate positions for 2017-18. Applicants should be recent Ph.D.’s with interest in DIMACS areas, such as computer science, discrete mathematics, statistics, physics, operations research, and their applications. There are four positions available:

  1. a one-year postdoctoral associateship investigating modeling of anomaly detection in multi-layer networks,
  2. a two-year associateship in collaboration with the Institute for Advanced Study (IAS) in Princeton, NJ emphasizing theoretical computer science and discrete mathematics,
  3. a position associated with the Simons Collaboration on Algorithms and Geometry which also emphasizes theoretical computer science and discrete mathematics and could be hosted at Rutgers/DIMACS,
  4. a two-year associateship in theoretical machine learning in the Department of Computer Science at Rutgers.

See the DIMACS website for application information.

Applications have various deadlines, beginning December 1, 2016. See website for details.
DIMACS Center, Rutgers University, 96 Frelinghuysen Road, Piscataway, NJ 08854-8018;
Tel: 848-445-5928; Email: postdoc at dimacs.rutgers.edu. DIMACS is an EO/AA employer.