NIPS 2017 Tutorial on Differential Privacy and Machine Learning

Kamalika and I gave a tutorial at NIPS last week on differential privacy and machine learning. We’ve posted the slides and references (updates still being made). It was a bit stressful to get everything put together in time, especially given how this semester went, but it was a good experience and now we have something to build on. It’s amazing how much research activity there has been in the last few years.

One thing that I struggled with a bit was the difference between a class lecture, a tutorial, and a survey. Tutorials sit between lectures and surveys: the goal is to be clear and cover the basics with simple examples, but also lay out something about what is going on in the field and where important future directions lie. It’s impossible to be comprehensive; we had to pick and choose different topics and papers to cover, and ended up barely mentioning large bodies of work. At the same time, it didn’t really make sense to put up a slide saying “here are references for all the things we’re not going to talk about.” If the intended audience is a person who has heard of differential privacy but hasn’t really studied it, or someone who has read this recent series of articles, then a list without much context is not much help. It seems impossible to even make a real survey now, unless you make the scope more narrow.

As for NIPS itself… I have to say that the rapid increase in size (8000 participants this year) made the conference feel a lot different. I had a hard time hearing/understanding for the short time I was there. Thankfully the talks were streamed/recorded so I can go back to catch what I missed.

Advertisements

ISIT 2018: call for CS theory papers too

I got an email from Venkat Guruswami encouraging those in the TCS community to submit work to the upcoming ISIT 2018 deadline. In particular, since ISIT papers are short (5 pages) it’s an ideal venue to publish more technical results or general tools (relevant to information theory) that get used in longer STOC/FOCS/SODA/etc papers. There was a lively discussion about what the “rules” were for ISIT, but basically:

  • the proceedings are archival so it counts as a real publication (no submitting the same result elsewhere)
  • ideal works would be things like coding theory problems of interest to both communities, TCS takes on IT problems, or general standalone results that could be applicable to information theory (or related) problems

The deadline is January 12, 2018. I guess I know what I’ll be doing for my winter vacation…

Rutgers ECE is Hiring (2018 edition)

My department is hiring for (potentially) multiple positions!

Hiring areas for this search are: (i) Electronics, including sensors, devices, bioelectronics, as well as integrated circuits and systems for RF and millimeter wave applications, (ii) Information processing and machine learning for autonomous systems and robots, especially learning and control in autonomous systems such as vehicles or drones as well as in assistive technologies, (iii) E-health, especially wearable electronics and sensors, medical informatics, quantified self and personalized medicine, as well as (iv) Cyber-physical systems, including signal processing and machine learning techniques, embedded systems, device and software security, IoT security, and applications to smart cities. Exceptional candidates in the university strategic areas are also welcome to apply.

NSF Report Markdown file

I experienced a horrible “network dropping causes web forms to clear” experience when filing an NSF report a few years back, so I switched to filling things in via the NSF’s Word template. However, the extraneous formatting in that made the cut-and-paste into the webworm tedious. So this time around I created a Markdown (.md) template with all of the questions you need to answer. This makes it easier to edit and lightly format your report text offline (e.g. on a plane) for much faster cut-and-paste later.

IPAM Workshop on Algorithmic Challenges in Protecting Privacy for Biomedical Data

IPAM is hosting a workshop on Algorithmic Challenges in Protecting Privacy for Biomedical Data” which will be held at IPAM from January 10-12, 2018.

The workshop will be attended by many junior as well as senior researchers with diverse backgrounds. We want to to encourage students or postdoctoral scholars who might be interested, to apply and/or register for this workshop.

I think it will be quite interesting and has the potential to spark a lot of interesting conversations around what we can and cannot do about privacy for medical data in general and genomic data in specific.

The Ideological Echo Chamber of the Beach

Disclaimer: I think that the beach should be open to all. Clearly there are some who unfairly malign some beach-goers. However, we need to have an honest discussion about the differences between different types of beach denizens. Thankfully, because I am a Star-Bellied and not Star-Eyed Sneech, I believe that we we can have an open and frank discussion about why Sneeches who originally did not have stars on their bellies have been disrupting the natural hierarchy of the beach. This bias towards equal-starness in Sneechdom threatens the viability of the beach and makes pariahs of those Sneeches whose came by their stars honestly [1]. It is imperative that we have this conversation now before the beach is destroyed forever.

There are obvious innate biological differences between Star-Bellied Sneeches and Common Sneeches. Star-bellied Sneeches have stars: this makes them more adept at skills that require people who have stars on their bellies [2]. Star-Belliedness is heritable: evolutionary sneechology shows that the reason Star-Bellied Sneeches are the “best kind of Sneetch on the beaches” because of the benefits conferred by their stars. What could be more simple? Of course, there are occasional bad Star-Bellied Sneeches and occasional good Common Sneeches, but the fact is on average, the Common Sneech is simply not as good on the beach as the Star-Bellied Sneech. Since the only difference between them is the star, it is only logical to assume that having a star makes a Sneech better at being on the beach.

Star-Bellied Sneeches are better at beach-related activities such as hiking and playing ball [3]. They are natural denizens of the warm beaches. By contrast, the Sneeches without stars are less adapted, preferring the “cold in the dark of the beaches” and “moping alone on the beaches.” This explains why the two types have different roles: Common Sneeches prefer the cold and moping alone because that is what they are good at. Unfortunately, because of our new culture of suppressing traditional viewpoints, we cannot have an honest discussion about this beach-appropriateness gap; the dire consequences of attempting to make all Sneeches have equivalent stars discriminates against the true Star-Bellied Sneeches. The time has come where a Star-Bellied Sneech can no longer speak of “frankfurter roasts, or picnics or parties or marshmallow toasts,” without requiring that Common Sneeches be invited too (despite the superficial appearance of their stars).

As an employee of Sylvester McMonkey McBean Corp. (SMMC), I feel it is my duty to speak out against unfair discrimination against Star-bellied Sneeches. The company offers special “fix-it-up” training and that puts stars on the bellies of Common Sneeches. What’s more, this star service is only available to Sneeches without stars! This is highly discriminatory: how can will we ever know whose stars are original and whose stars are the result of special training? This violates the natural order of things. Because Common Sneeches have a tendency to mope alone, we run the risk of having Sneeches without innate hiking- and ball-playing-skill gaining prominence on the beech. Again, this is not bad for specific Sneeches, but on average will make SMMC less competitive and the beach sub-optimal.

My recommendation is that we should treat each Sneech individually. Those Common Sneeches who are demonstrably non-mopey and can, in controlled environments, enjoy a marshmallow toast, could be possibly considered for further admission into frankfurter roasts and picnics. However, we cannot simply erase differences between Sneeches by allowing non-Star-Bellied Sneeches to obtain stars. It may be politically incorrect to say this, but we must have a way of distinguishing those Common Sneeches who are deserving of stars from those who are not [4].

Sylvester McMonkey McBean Corp. is treading down a dangerous path and is alienating a great number of Star-Bellied Sneeches. I have received very positive feedback regarding these opinions, so there is clearly a “silent minority” of Star-Bellied Sneeches who are yearning for an open conversation [5] about these topics. However, the beach has become an inhospitable plate for the rational-minded Sneech.

In closing, I wish to state here that I do not intend to offend Common Sneeches: therefore I cannot possibly have offended any of them, because I said so. That is logic that any Star-Bellied Sneech can understand: Sneeches who cannot are clearly Common mopes.

[1] That is, they were born with them.

[2] For example, fooling starfish while snorkeling is something only Star-bellied Sneeches can do.

[3] While it’s true that in previous years “you could only play ball if your bellies had stars,” I would like to strenuously emphasize that that is #NotAllStarBellies. Not mentioning this disclaimer is evidence of the ideological commitment to the notion that all Sneeches are similarly capable of playing ball. Common Sneeches should perhaps play in a league of their own.

[4] Although there may be Star-Bellied Sneeches who fail to meet the bar, this can be addressed by introducing better role models and mentors. Unfortunately, such programs are only available to Common Sneeches.

[5] Preferably one only involving non-mopey Sneeches.

100 signatures needed by 5/12 to nominate José Moura for IEEE President

As I wrote earlierJosé Moura is trying to get on the ballot for IEEE President. The deadline is May 12 and he needs only 100 more signatures to get onto the ballot. If you are an IEEE member please take the 1 minute to sign the petition to get him on the ballot.

IEEE has an undemocratic nominations process in which the IEEE Board of Directors (BoD) gets to decide who the candidates will be. Because Prof. Moura opposed the BoD proposed amendment to consolidate power into the BoD and reduce regional representation, it is not hard to imagine that the BoD would not want to allow dissenting voices in the Presidential race. 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.”

Put José Moura on the ballot!