Skip to content

Nina Balcan Wins

April 10, 2020


Congrats and More

[ CMU ]

Nina Balcan is a leading researcher in the theory of machine learning. Nina is at Carnegie-Mellon and was previously at Georgia Tech—it was a major loss to have her leave Tech.

Today we applaud her winning the ACM Hopper Award.

ACM President Cherri Pancake says:

Although she is still in the early stages of her career, she has already established herself as the world leader in the theory of how AI systems can learn with limited supervision. More broadly, her work has realigned the foundations of machine learning, and consequently ushered in many new applications that have brought about leapfrog advances in this exciting area of artificial intelligence.

The Hopper Award

The ACM Grace Murray Hopper Award is given to: An outstanding young computer professional, on the basis of a single major contribution before the age of 35. Here are five of the most recent winners:

  • Constantinos Daskalakis, (2018)

  • Michael Freedman, (2018)

  • Amanda Randles, (2017)

  • Jeffrey Heer, (2016)

  • Brent Waters, (2015)

Nina’s Contribution

The Hopper award says it is for a “single” major contribution. I believe that Nina is almost a disproof of this statement: I fail to see how she only did one major contribution. In fact, the citation lists three. In any event I thought we might look at one of her top results on learning. It is a paper from 2006 with over 400 citations. The ttile is “Agnostic Active Learning” and is joint with Alina Beygelzimer and John Langford.

Active learning follows a classic idea in computer theory: Making a protocol interactive can often decrease the cost, and almost always makes the protocol more complex to understand. In active learning one is given unlabeled examples. As usual the goal is to classify the samples. However, as the samples are unlabelled, the learning can ask for labels for elected samples—this is the active part of the learning. As you might imagine asking for labels has a cost, so the learner strives to ask for the fewest labels possible.

The savings can be large when the labels are perfect—that is, noise-free. In general it is much more complex to understand when active learning helps. Nina’s work found examples where noise can be tamed. Her award citation says:

Balcan established performance guarantees for active learning that hold even in challenging cases when “noise” is present in the data. These guarantees hold under arbitrary forms of noise, that is, anything that distorts or corrupts the data. This can include anything from a blurry photo, a unit of data that is improperly labeled, meaningless information, or data that the algorithm cannot interpret.

See her papers for the details.

Other Awards

There are various awards for computer scientists, many are from the ACM. Since Alan Perlis won the first Turing award, there have been 69 more winners. Only three have been women:

  • Shafi Goldwasser, (2012)

  • Barbara Liskov, (2008)

  • Frances Allen, (2006)

Here is another quote from the president of the ACM:

We typically receive one woman nominee [for the Turing Award] every five years. It’s very disturbing.

The number of nominations is too small. There are plenty of strong women candidates for the Turing award, and for other awards. We need to do a better job. See this for more thoughts on this issue.

Open Problems

We do not know about the situation with nominations for the Hopper Award. Nina is only the seventh woman to win since 1971, but four of the last ten Hopper Award winners have been women. How can we recognize more women under 35 who are doing great work?

Again we congratulate Nina Balcan on her richly deserved honor.

[Fixed typo “Congrats” ]

7 Comments leave one →
  1. April 10, 2020 10:34 pm

    ML is still mystic to me. I do not know what counts as important and what counts as mediocre. Not clean cut as cstheory.

    • April 10, 2020 10:36 pm

      Why is her work considered important? Could any ML amateur couldn’t have done it? It is not as clean cut as your other theory promotions.

  2. April 11, 2020 10:06 pm

    Comments were blocked.

    • April 12, 2020 9:44 am

      First comment is moderated, thereafter comments go through freely. We have been diverted by another story overnight.

      • April 13, 2020 2:08 am

        Why is her work considered important? Could any ML amateur couldn’t have done it? It is not as clean cut as your other theory promotions. “Balcan established performance guarantees for active learning that hold even in challenging cases when “noise” is present in the data. These guarantees hold under arbitrary forms of noise, that is, anything that distorts or corrupts the data. This can include anything from a blurry photo, a unit of data that is improperly labeled, meaningless information, or data that the algorithm cannot interpret.” looks like standard type of result in communication complexity. What qualifies as ground breaking in ML and what does it even mean to be fundamental there? And is fundamental work any important that work done at google and fb?

    • April 14, 2020 7:52 am

      Well again.

  3. April 13, 2020 6:27 pm

    In her paper her name is Maria-Florina, not Nina

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s