Michel Talagrand wins Abel Prize for work wrangling randomness | Hacker News Reader

Michel Talagrand wins Abel Prize for work wrangling randomness

quantamagazine.org

278 pointsjnord48 comments

What an amazing output.

Reading his papers, he manages to connect basic geometric ideas (like the notion of "isoperimetry", i.e., enclosing the maximum volume within a given perimeter) to probabilistic notions like convergence of averages. It links together stuff you learn in information theory (the probability mass of the "typical set"), and in machine learning (deviations of an training-set error rate from a true error rate), and in probability theory (empirical processes).

His papers typically had an introduction that related the main theorem to some of these basic geometric notions. The introduction would reveal a whole new, very intuitive and geometric connection between a very abstract theorem to basic geometry, like the volume of a spherical shell. It would routinely blow my mind.

I really enjoy this sort of story. Someone persevering in an unpopular arena for the love of it, and then having it become quite essential for the advancement of humanities goals as it is developed.

thats pretty much also the story of deep learning isn't it, i too enjoy reading about researchers who have a pure devotion to their subject.

This is such a great quote for everyone! No matter the age. No matter what one wants to do.

> “I’m not able to learn mathematics easily,” Talagrand tells ... “I have to work. It takes a very long time and I have a terrible memory. I forget things. So I try to work, despite handicaps, and the way I worked was trying to understand really well the simple things. Really, really well, in complete detail. And that turned out to be a successful approach.”

Just imagine. You may be super smart who gets things easily and right away. Or, you may be average. Using this philosophy in life, one can excel further.

I didn't know about his inequalities, but I found (1) that provides an example of using Talagrand inequality applied to the longest increasing subsequence problem (12 pages, easy to read). It seems to be a broad generalization of the Hamming distance.

(1) https://www.cs.purdue.edu/homes/hmaji/teaching/Spring%202018...

I was about to say that. Also, check out this

http://doc.9gridchan.info/blog/181230.ncubic.routing http://doc.9gridchan.info/blog/190104.ncubic.algorithms

Fromm mycrovtif (RIP), a guy from the plan9/9front community.

> fell in love with his future wife, a statistician, at first sight (he proposed to her three days after meeting her)

Was that after sampling 1/e prospects?

(The "marriage problem" is applied statistics on when it's optimal to stop searching for something https://en.wikipedia.org/wiki/Secretary_problem )

Following Shaw Price in 2019 https://invidious.projectsegfau.lt/watch?v=22hR2CkNGXE

I've used and read about many concentration inequalities in machine learning and other applied math disciplines (Hoeffdings, McDiarmid, etc).

Talagrands' results seem to generalize those, but I haven't had the chance to see them in the wild (yet).

> He recalls with delight that he once used a cab service whose owner recognized his name, having learned the inequality during a probability class in business school.

The taxi drivers in France are cut from different cloth!

As well as their business students, apparently.

The French higher education system is such that a lot of business students come from somewhat math-heavy undergraduate programs: 9h hours of math class plus one hour of one-on-one oral examination for 2-3 years.

I think you missed the part about the owner was the one who went to business school.

I know Talagrand because some of his work comes up in topological dynamics (work around Rosenthal's l¹-dichotomy culminating in the Bourgain-Fremlin-Talagrand dicothomy for compact sets of Baire class 1 functions), but I had no idea he has such accomplishments in other fields! Impressive

Slightly off-topic - has anyone read his book "What is a Quantum Field Theory"? Can you recommend it?

See a review here: https://www.math.columbia.edu/~woit/wordpress/?p=12819

Thanks! I just saw Peter Woit's post on the Abel Prize https://www.math.columbia.edu/~woit/wordpress/?p=13877 also speaks positively about that book.

I've been reading it over the past few days (picked it up after seeing it mentioned here). It's fantastic, much better than many similar attempts to make quantum field theory comprehensible to mathematicians.

In the NewScientist interview he said, I tried “to understand really well the simple things.”

Ironically, the Wikipedia pages about him in French [1] and English [2] are, currently, super short. It will help if Wikipedia (just giving them as an example) could handle the focus on certain people and topics in an smarter way and not after the facts. Wikipedia has a special opportunity about handling this because it is the canonical general reference on Internet.

Hallucinated or not I found a better starting point on [3], a Reddit post of 2 years ago with a comment saying:

"Even professional mathematicians are barely qualified to choose the best mathematician in their narrow field of expertise, let alone in general…

(Before I left math, the most difficult work I encountered was by Michel Talagrand.)" [4] and last, but would be first indeed his own web page [5]. He even gives prizes ala Knuth for solving specific math problems.

Last, really few mentions in *stackexchange.com and Reddit.

[1] https://fr.wikipedia.org/wiki/Michel_Talagrand

[2] https://en.wikipedia.org/wiki/Michel_Talagrand

[3] https://chat.openai.com/share/39374448-da85-4897-977a-aaa37e...

[4] https://www.reddit.com/r/math/comments/s81ysm/who_would_you_...

[5] https://michel.talagrand.net/

[6] https://www.google.com/search?q=Michel+Talagrand+site%3Astac... and https://www.google.com/search?q=Michel+Talagrand+site%3Aredd...

That was me on Reddit. I emailed him asking about some of his work on invariant means in the 1970s. He said “I had no taste then”, and told me what a waste of time it was.

At that point, I decided to go into data science instead of trying to get a post doc…

How do you feel about that decision now? I left math after my PhD and often find myself feeling nostalgic about it.

Always something to look forward to in [early] retirement to keep the mind active.

> [...] and not after the facts

Wikipedia is an encyclopedia, they intentionally only cover topics retroactively, and preferably after the dust has settled. They're intentionally not intending to be many things, including a source of up to date news: https://en.wikipedia.org/wiki/Wikipedia:What_Wikipedia_is_no...

No pun intended. We all see (right?) that one of the cultural topics of the XXI is about focus. In a relative past people could study and/or read "all" texts and be intellectuals or feel like that. It is obvious that the focus economy always existed but now there are zillions of potential content to consume (good or not, doesn't matter).

Wikipedia, as an example, has the opportunity to add layers (not many) of content in the quest of helping (not solving) this focus problem. As you said Wikipedia does not currently has this purpose, but this does not mean that they cannot carry the lit torch and pay attention to the focus economy (wordplay just by chance).

It is also important to highlight that Wikipedia has many externalities, it is not just them. For example, Wikipedia results are generally the first that cames up in a search engine and their content is much used in machine learning. In this context, the problem of focus is not just about Wikipedia itself but the "focus graph" that has Wikipedia as one of the top releveant nodes.

Wikipedia is premised on being a tertiary source, and pays a price for every "layer" of content they add in volunteer time needed to correct inaccuracies, which are constantly being added. People have all sorts of ideas about things Wikipedia could be besides an encyclopedia, but the project is (wisely) not receptive to any of them. Meanwhile: if you're right about this, you can just fork and track Wikipedia and do your own site.

Even, beyond that, Wikipedia, again, as an example, has faults on its own merits: I keep reading about Michel Talagrand and find this entry [1] which mentions his work first but there is no link in [2] to [1] (at least now).

I will try to restate this, for the sake of an interesting discussion, in a completely different direction but using Wikipedia as an example: in software engineering we create different kind of tests for our software systems, I think Wikipedia should add "unit tests" and other tests to augment, fix, and link their current content.

[1] https://en.wikipedia.org/wiki/Majorization

[2] https://en.wikipedia.org/wiki/Michel_Talagrand

Wikipedia has more process (and is by most measures, more important and successful) than most software most people work on so I'm not sure tips from software engineering is what they are missing.

Are you saying that Wikipedia doesn't need to innovate and just leave their organization as it is? I follow Wikipedia since its inception and it seems part of its processes, as in major organizations, amplifies bureaucracy more than process innovation. Any organization is not only their throughput but their processes, including resources as humans.

I'm sure there's any number of things they can do better, but revisiting their entire premise isn't just a suggestion, it's a request that they become a completely different thing.

Wikipedia is probably the single most successful human knowledge project of the last 100 years. It sounds crazy saying that out loud! Maybe it's not true! But that it's even a colorable argument speaks to how little software engineers have to contribute to its fundamental direction. It's not about us.

Last 5000 years.

Personally I don't care about AI, but there would be no AI without Wikipedia.

I agree with you with the importance of Wikipedia but it should be noted the acceleration of technology (not talking about singularity!) that will impact them. It is the acceleration of the means in the relationship of humans with technology. This is a fact.

The innovation dilemma is always present, even for NGOs.

The point is that his work (and by extension himself) was encyclopedia-worthy long before he got the prize. Or he wouldn't have gotten the prize. Wikipedia doesn't wait until after the oscars to include movie.

All of us were always free, and are still free, to contribute to making his wiki page better!

Wikipedia editors delete pages about people not deemed notable enough. Especially women: https://www.cbc.ca/news/science/wikipedia-bias-1.6129073 Meanwhile, a list of supplemental roads and rural secondary highways in Kentuck (3000-3499) is encyclopedia-worthy: https://en.wikipedia.org/wiki/List_of_Kentucky_supplemental_...

If we assume that there is a group of people giving a concerted effort to acknowledge women in general, would that not lead to more articles about women being created? And, purely statistically, would not this in turn lead to more articles about women being rejected for being insufficiently notable? There is no need to assume bias in Wikipedia editors, the bias could just as well lie with those people who create the many rejected articles.

Also, one criteria which Wikipedia must, by necessity, use, is “published articles in other media about the subject”. If other media are, in general, biased, this would lead to a dearth of articles in other media, which in turn would lead to Wikipedia rejecting new WP articles.

What evidence is there for said concerted effort?

<https://en.wikipedia.org/w/index.php?title=Wikipedia:WikiPro...>

How is this statistical problem (which I’m prepared to acknowledge) an actual barrier preventing you from improving any given article?

Because I'm not going to put in the work just to see it deleted by zealous editors.

There is no basis in your purported belief that quality contributions that follow the rules will get deleted.

Wikipedia editors are not out there to suppress women— The fact that this de facto happens is more likely a reflection of systemic social issues.

Surely Wikipedia editors are part of society? So

> is more likely a reflection of systemic social issues.

contradicts

> Wikipedia editors are not out there to suppress women

The basis for my belief that it happens is that I've seen it happen.

Links or it didn’t¹.

1. <https://news.ycombinator.com/item?id=36123641>

Here is some starting point for your search: https://www.cbc.ca/news/science/wikipedia-bias-1.6129073 See also: https://en.wikipedia.org/wiki/Deletionism_and_inclusionism_i...

I wanted you to provide links to what you have seen happen, like you said you had.

His homepage has this interesting snippet:

> If you are desperate to get my books and your library can't afford them, try to type the words "library genesis" in a search engine. I disagree with piracy, but this site saved me many trips to the library, which unfortunately does not carry electronic versions of older books.

See https://news.ycombinator.com/item?id=18131009 for an earlier discussion of scientific blind spots in Wikipedia.

I also wish there was a mechanism to prune certain Wikipedia pages that carry way too much detail, given the notability of their subject.

Reply on news.ycombinator.com