Slashdot discussion on “High Paying Jobs in Math and Science”

Slashdot is having a very active discussion (naturally degenerating into nonsense at times, but generally pretty good) about high paying jobs for those will a college degree in math or science. Operations research gets a good plug or two among the discussants. One response (from “MoneyCityManiac”):

Applied math is a good bet. Operations Research (“OR”), as Wikipedia defines it, is “an interdisciplinary science which uses scientific methods like mathematical modeling, statistics, and algorithms to decision making in complex real-world problems which are concerned with coordination and execution of the operations within an organization.” It’s a mixture of math, stats, CS, and engineering.

There’s OR applications in areas such as health-care, environmental management, forestry management, transportation, and much more. Environmental management, in particular, is something that operations research is going to play a huge role as government and industry focus on reducing greenhouse gas emissions.

And because there’s such a practical role towards it, there’s plenty of support from government and industry, not just in terms of jobs at the end but also scholarships, fellowships, etc. Ask around a math, CS, or engineering department! I’m sure it won’t be hard to find someone who can point you in the right direction.

Operations Research in the New York Times and I am annoyed and depressed

The term “Operations Research” appears in the New York Times today (May 20, 2007) in an article entitled “Reaping Results: Data-Mining Goes Mainstream“. Here is the start:

RODNEY MONROE, the police chief in Richmond, Va., describes himself as a lifelong cop whose expertise is in fighting street crime, not in software. His own Web browsing, he says, mostly involves checking golf scores.

But shortly after he became chief in 2005, a crime analyst who had retired from the force convinced him to try some clever software. The programs cull through information that the department already collects, like “911” and police reports, but add new streams of data — about neighborhood demographics and payday schedules, for example, or about weather, traffic patterns and sports events — to try to predict where crimes might occur.

Later comes the “Operations Research” reference:

The desire to exploit computing and mathematical analytics is by no means new. In the 1960s and ’70s, “operations research” combined computing and math mainly to make factory production work more efficient. And in that period, “decision support” software was intended to help managers more intelligently use information in the big computing file cabinets — databases — that were becoming common in corporations.

That really burns my butt. I hate the idea that operations research is shunted into history and it is the new analytics that is the key. It is all Operations Research!

Otherwise, the article is great. Analytics are a key success path for companies. Here is a quote from Wired on why Yahoo! lost out to Google:

At Yahoo, the marketers rule, and at Google the engineers rule. And for that, Yahoo is finally paying the price.

On the personal side, I also cringed when I saw the following in the New York Times article:

“It’s really starting to become mainstream,” says Mr. Davenport, co-author with Jeanne G. Harris of “Competing on Analytics: The New Science of Winning” (Harvard Business School Press, 2007). The entry barrier, he says, “is no longer technology, but whether you have executives who understand this.”

This was really the theme of my Hood Lecture, and I had the inklings of writing something longish on the subject. I will have to see if this book does a good job on this.

OR, Poker, and Teaching

It is a lovely morning here in New Zealand, and the sun is rising over the bay that my house overlooks. So naturally I am wandering around the web.

Gary Carson’s Math and Poker blog is one that I regularly follow (not the least because he points to this blog). He writes about the role OR plays in understanding poker and about how OR is taught:

Part of the problem that operations research has in getting recognition is the way we teach it — it’s taught as a bunch of algorithms for the most part. Even when it’s taught as a bunch of models to be applied to real problems, the models are taught as members of a tool kit. Seldom do we teach OR as a process of using models to abstract or isolate the elements of a problem that are critical to decision making related to that problem.

I think OR education needs to put more of a focus on using the model and less focus on solving the model. I think students would form a deeper grasp of what OR is all about if that was done. Stuff like analysis of residuals and sensitivitey of LP solutions are just too important to be glossed over.

I teach at a business school and we have moved much more to teaching about using models for real-world business decisions. Things like sensitivity do end up taking a backseat, however, since without understanding the underlying algorithm/mathematics things like dual values are just mysterious black boxes. The challenge is, I think, how to get enough of the fundamentals across so people can confidently and correctly use the resulting models. And I think we are all over the map on this at the moment, to the detriment of the field.

Quantum Computing and Learning from Blogs

Scott AaronsonEver since I heard about quantum computing (it arose in the 1970s, but most of the really visible stuff started in the mid 1980s, see the Wiki timeline), I have been skeptical. This skepticism arose from a view that quantum computers could solve NP-complete problems in polynomial time. I was randomly wandering the blogosphere when I can across “Shtetl-Optimized” by Scott Aaronson, with the subtitle “Quantum computers are not known to be able to solve NP-complete problems in polynomial time”. Well, perhaps everyone else knew that, but it was somewhat annoying that something I knew for two decades was dead wrong. While practical quantum computers still seem some time away (it seems the state of the art is factoring 15=3×5), trying to determine the effect of quantum computing on OR seems like an interesting question.

One post of Scott’s that I like very much is his “The Hiring Scott Aaronson FAQ“, where he lists some of the questions he was asked in his job search (he is a postdoc at my alma mater, the University of Waterloo’s Department of Combinatorics and Optimization). I’ve interviewed for a job or two in the couple years this blog has been going, but I have not been bold enough to talk about it. Scott uses the entry as a very quick survey of what he believes in, and the interviewers don’t come across like idiots (I guess I would have asked about half the questions myself).

Check out also his article “NP-Complete Problems and Physical Reality“published a couple of years ago in the SIGACT News complexity column. Having lived through an era when NP-completeness results were being churned out by the boatload with only a few providing real insight and advance (kinda like approximation results these days), I have not advanced very much beyond what I knew ten years ago, but Scott’s writing make this look pretty interesting again.

Finally, I am jealous of Scott’s ability to write well enough and intriguingly enough to generate dozens of comments on each of his posts. He points to the following review of his blog:

OK, on to the second breakfast link. Bill Gasarch has reviewed my blog for SIGACT News (scroll down to page 15), together with Lance Fortnow’s and Luca Trevisan’s. Favorite quotes:

Lance is Walter Cronkite. Scott is Stephen Colbert.

The name of the blog, ‘Shtetl-Optimized’ does not really say what it is about. With this in mind one can never say Scott has gone ‘off topic’ since its [sic] not clear what the topic should be.

Perhaps I am sticking to the topic too much!

Videos and Operations Research

Dick Larson has a very nice editorial in the April 2007 OR/MS Today on the use of videos in disseminating information about our field. One big point he makes is that the Edelman videos, a tremendous resource for our field, are not easily and freely available. This frustrates the heck out of me too! I do not understand why videos of the contestants are not up on YouTube right after the competition. There are some issues of ownership of the various INFORMS entities, but some INFORMS President will eventually be able to work them out and suddenly the world will be able to see how amazing we are.

Until then, Larson has another article on MIT World, where MIT professors give lectures on various topics. Arnie Barnett and Yossi Sheffi, along with Larson, are among those giving OR-oriented talks.

Incidentally, there are some “Operations Research” videos at YouTube but only 8 at the moment. Let’s see if things improve in the next year.

Operations Research and Turing Machines

From slashdot.com:

An anonymous reader writes “Stephen Wolfram, creator of Mathematica and author of A New Kind of Science, is offering a prize of $25K to anyone who can prove or disprove his conjecture that a particular 2-state, 3-color Turing machine is universal. If true, it would be the simplest universal TM, and possibly the simplest universal computational system. The announcement comes on the 5-year anniversary of the publication of NKS, where among other things Wolfram introduced the current reigning TM champion — ‘rule 110,’ with 2 states and 5 colors.”

Operations research (particularly integer programming) seems relevant to this work (through things like the undecidability of integer quadratic programming). $25,000 anyone?

Operations Research and Wikipedia

As part of my overview talk, I took a look at Operations Research in Wikipedia. Seeing a lack of pointer to this blog (which I think is a pretty good OR blog!), I went to fix it, but decided I had a Conflict of Interest, so reverted. I’ll leave it to others to decide if it should be added.

Anyway, I have mixed feelings about the OR section in Wikipedia. The long section on World War II applications simply reinforces the absolutely incorrect belief that OR may have been useful historically but has been superseded by newer, flashier methods. They are great stories, but they don’t really reflect the reality of OR today. I would much rather have summaries of recent Edelman finalists.

Some of this may be more general problems of describing science on Wikipedia, as this Wired blog entry by Thomas Goetz discusses:

Wikipedia is, by all measures, one of the great accomplishments of the Internet Age. I’m willing to say it stands alongside Google, eBay, GoogleMaps, IMDB and Wired.com as among the greatest resources on the Web (ok, that last one is self-serving).

But boy, does it suck when it comes to science topics.

Of course, there is nothing to stop me from doing some editing of the OR entry(short of COI), but I am trying to cut down on things I am doing. Perhaps someone will take it on as a weekend project. And I think an OR-oriented Wiki would be a great project.

Interaction of Computers and People

In a few days, I give a “public lecture” as part of the requirements for my Hood Fellowship that the University of Auckland gave me (I would have visited here anyway – New Zealand is wonderful – but the Fellowship was a very nice addition). Rather than trot out a version of my “sports scheduling” talk, I am giving an overview of OR talk. It is really quite a challenge. These overview talks can be quite a snooze for those in the audience who know OR, so I am thinking hard about the role OR has in the world, and the challenges the field faces. The “Science of Better” initiative of INFORMS has been very helpful in this regard.

One aspect I am exploring is the appropriate role of OR in decision making. While optimizers like me tend to be a bit impatient with this (“The best answer is this! Stop arguing with me and do it!!), the real-world is full of stuff that doesn’t appear in our models. In many (most?) applications, OR is an adjunct to decision making, not a replacement. Slate has an interesting article about this in the context of chess-playing programs that I may try to work into my talk.

Apologies in advance: the next few posts may be me trying to work through my thoughts in preparation for the talk.

Mathematical Puzzles, Martin Gardner, and Peter Winkler

I am certainly not alone when I say that interest in mathematics was sparked by Martin Gardner’s Mathematical Games column in Scientific American. I have a strong memory of many boring physics classes in high school which I whiled away reading through the stack of Scientific Americans in the corner. Those columns led to mathematics in university (where I probably not coincidently received a “D” in physics) and onward through my interest in discrete optimization.

While Gardner has long since stopped writing the column, his influence remains strong. Every year, a group of mathemeticians, magicians, jugglers, puzzle makers and so on meet at a “Gathering for Gardner“, a get-together that looks to be a blast!

Last year, Peter Winkler, formerly of Lucent and now at Darmouth, put together a set of puzzles, cleverly and correctly entitled “7 Puzzles You Think You Must Not Have Heard Correctly”. Here is one I particularly like:

The names of 100 prisoners are placed in 100 wooden boxes, one name to a box, and the boxes are lined up on a table in a room. One by one, the prisoners are led into the room; each may look in at most 50 boxes, but must leave the room exactly as he found it and is permitted no further communication with the others.

The prisoners have a chance to plot their strategy in advance, and they are going to need it, because unless every single prisoner finds his own name all will subsequently be executed. Find a strategy for them which which has probability of success exceeding 30%. Comment: If each prisoner examines a random set of 50 boxes, their probability of survival is an unenviable 1/2^100 = 000000000000000000000000000008. They could do worse if they all look in the same 50 boxes, their chances drop to zero. 30% seems ridiculously out of reach but yes, you heard the problem correctly.

Is this really possible? Check out Peter’s writeup for the solution (and seven other problems). Also check out his book for more.