Blogs and Labs

Why don’t more research labs have blogs? Most of us are passionate about our fields and our perspective on the world. For instance, I really believe operations research is a great way to view the world, and I trust that comes out through the blog.

David Goldberg of the Department of Industrial and Enterprise Systems Engineering at the University of Illinois is another with a passionate view of how his lab’s research fits into the world, and it comes through in the lab blog. I even get a mention as “worth a read”. David is, of course, a guru on genetic algorithms, an area I consider part of operations research, though I suspect David disagrees. In any case, we need more discussion and even conflict if our field is to be taken seriously.

Another lab with a blog is at the Missouri Estimation of Distribution Algorithms Lab.  To be honest, I did not know what an “Estimation of Distribution Algorithm” (EDA) was.  I do now, due to the blog.

The BBC and queues

The BBC has an economics editor named Evan Davis who has a blog on understanding the real world using the tool kit of economics. There are two recent entries (here and here) on queueing results. Now the English are famous for their queueing discipline, but Davis is concerned with two examples of “non-obvious” queues: road congestion and waits for health services. In both cases, there are significant costs to queueing and the economic effect is that since people enter the queue only when the benefit of the service outways the cost of queueing, many people end up with very little value (the queue increases to wipe away the consumer surplus). This is a nice example of combining economics with operations research (many OR models have very simple queue entry rules, like “will enter always” or “will enter only if queue size <=12"), but using economics to guide the consumer behavior leads to better insights. Davis shows his economist (not OR!) training in stating the following:

There are three very simple rules about queues.

• 1. They grow longer when the number of people joining at the back of the queue exceeds the rate at which people are being dealt with at the front.

• 2. They grow shorter when the rate at which people are dealt with at the front exceeds the rate at which people join at the back.

• 3. They stay constant when the flow of new arrivals is equal to the flow of people being seen.

It is the last point that is incorrect, assuming there is any variation at all in either service or queue entry. If flow of arrivals equals flow of people being seen, the queue will continue to increase (on average). This point is actually very important. Managers are constantly being told to look for inefficiencies, and having more service than customers looks like an inefficiency, so they work at getting service = customers. The result is long queues (unless there is no variation in service or queue entry). Unless there is slack in the system, the system doesn’t work.

Davis does mention Operations Research (surprisingly: Operational Research would be more “british”):

Queuing theory is in fact, quite a science. It comes up in the discipline of Operations Research, which studies processes, production and organisations using maths, statistics, economics and management science. It’s a fascinating subject.

The “It’s” in the last sentance is unclear. I’ll take it to mean that “Operations Research is a fascinating subject”, which is certainly the case.

MIT Conference on Sports Business

MIT is holding a conference this weekend on Sports Business (unfortunately it is sold out) with a focus on Analytical Sports Management. This workshop is an interesting mix of sports insiders, economists, operations researchers, media executives and more, with a strong emphasis on those in the business. No talks directly on scheduling (my particular emphasis) but a cool looking conference nonetheless.

Brenda Dietrich in Fast Company

Brenda Dietrich,  President of INFORMS and head of Math Sciences at IBM Watson Research is profiled in Fast Company this month.  Some wonderful stories:

If you’re not a mathematician, the deep math that Dietrich and her team perform sounds utterly foreign–combinatorial auctions, integer programming, conditional logic, and so on. Their whiteboard scribbles at Watson look incomprehensible, like Farsi or Greek (then again, many of the symbols are Greek). But these mysterious equations represent the real world and how it works. When mathematicians “model” a problem, they’re creating a numerical snapshot of a dynamic system and its variables.

Take the forest-fire project Dietrich and the researchers are working on. Extinguishing fast-spreading flames over tens of thousands of acres is an expensive and complicated undertaking. In 2000, a particularly devastating year, the federal government spent more than $1 billion and still lost more then 8 million acres. Its fire planners want to reduce the cost and the damage through better coordination among the five agencies involved.

Armed with seven years of data, IBM’s mathematicians are creating an enormous model that shows how the resources–every firefighter, truck, plane, etc.–have been used in the past, how much each effort cost, and how many acres burned. The algorithms describe the likely costs and results for any number of strategies to combat a given fire. “How many bulldozers and buckets do you keep in Yellowstone Park?” Dietrich asks. “And if you need to move them elsewhere, how much will it cost and how long will it take?” She’s talking fast, describing the unruly variables that math makes sense of. “It’s a nice project. Complicated, huh?”

It is too bad that Brenda is described as a mathematician (which she is) rather than the more specific and accurate “Operations Researcher”.

Do Operations Research, win $1 million

Art Geoffrion wrote me, pointing out that the Netflix Prize is a great opportunity for OR people to show their stuff. Netflix is offering up to $1 million for a system that predicts whether a customer will like a movie or not. They have made available a wonderful database of 100,000 ratings. Lots of people have used data mining methods on this database For me, the line between data mining and OR is very thin indeed, so it would be interesting to see what an OR approach can do with this.

The Wall Street Journal has an article on these types of prizes. There are a lot of good reasons for companies to provide these competitions:

Prizes prompt a lot of effort, far more than any sponsor could devote itself, but they generally pay only for success. That’s “an important piece of shifting risk from inside the walls of the company and moving it out to the solver community,” says Jill Panetta, InnoCentive’s chief scientific officer. Competitors for the $10 million prize for the space vehicle spent 10 times that amount trying to win it.

Contests also are a mechanism to tap scientific knowledge that’s widely dispersed geographically, and not always in obvious places. Since posting its algorithm bounty in October, Netflix has drawn 15,000 entrants from 126 countries. The leading team is from Budapest University of Technology and Economics.

Given the generality of OR, it is clear that our field can be competitive in many of these. Any takers?