Month: July 2010

Another Operations Research Dean

Sunil Kumar, currently at Stanford, has been named Dean of the University of Chicago Booth School of Business.  Sunil’s research has been in manufacturing, stochastic control, queueing networks, and related areas, making him clearly an operations research person (and of course he is a member of INFORMS).  He is an area editor of Operations Research, a position I suspect he will be giving up shortly.

It is amazing how often the field of operations research leads to academic administration.  Congrats Sunil!

A Kiwi on the Move

I spent 2007 visiting the University of Auckland, living on the wonderful (and strange) island of Waiheke, and I have about a thousand pictures online to prove it.  We had a wonderful year, and loved the university, the island, the city, and the country.  So it is with some jealousy that I note that Tava Olsen, formerly of Washington University, St. Louis, has taken up a position at the University of Auckland as the Ports of Auckland Chair in Logistics and Supply Chain Management.

Tava’s family also lived on Waiheke and we visited with her on the island a couple times during the year we were there.  David Ryan, who has done great work in airline scheduling and branch-and-price, has a place on the island.  Garrett van Ryzin lived on the island during his sabbatical year.  With just under 8,000 people, Waiheke has to have one of the larger “output of operations research per capita” values in the world!

I think to celebrate her new Chair, Tava should invite all of us to Waiheke for a nice glass of wine.  Congratulations, Tava!

Eating Better and Better Routing

For the last year or so, my wife and I have decided to eat better by doing more “real” cooking.  A great help in this has been a magazine “Real Simple“.  Every month, the magazine publishes a series of recipes, each generally requiring only 20-30 minutes of preparation time.  We like these recipes because they use real ingredients:  none of this “Pour a can of cream of celery soup over everything”.  We’ve agreed to cook everything, whether it sounds appealing or not, and of the dozens of recipes we have gone through, essentially all of them were edible, with most very good and a few definite keepers (*).   The authors of the recipes do seem to have a fondness for leeks and fennel, but we have grown used to that.   Alexander, my six year old son, eats the same food of course, and generally approves of what we are cooking.

I was delighted with this month’s issue where they had a short blurb on the website route4me.com.  The description appeals to their readership:

You need to get to the library before closing, but you also have to pick up the dry cleaning, the kids from school (don’t forget that one), and the inevitable snack along the way.  Enter up to 10 addresses on this site and it will calculate the shortest route to get it all done, complete with driving directions.

The Traveling Salesman Problem makes an appearance in our cooking magazine!  Naturally I went over to the site, and checked it out by putting in a few cities (seems a limit of 6 but maybe signing up gets you more): Pittsburgh, Cleveland, Atlanta, Winnipeg, Minot (ND), and Los Angeles.  Clicked “round trip” to get back home and ended up … with a somewhat surprising route:

Hmm… that crossing in Illinois is a little suspicious.  This doesn’t look like the optimal route.  Is it? Maybe it is hard to get from Cleveland to Winnipeg due to the lakes?  Perhaps here was an example were the underlying road network really has a strong effect on the optimal tour.

I checked things out, and compared this route to the route going from Pittsburgh-Cleveland-Winnipeg-Minot-LA-Atlanta-Pittsburgh.  Turns out the route from route4me is about 10 hours (driving) longer than the crossing-free route.  What kind of optimization is this?

It took a bit more playing around before I figured out what route4me was doing.  Their definition of a “round trip” is the minimum path visiting all the cities from the starting point, followed by going from the final city back to the starting point.    The best path is Pittsburgh-Cleveland-Atlanta-Winnipeg-Minot-LA;  they then just add in the LA-Pittsburgh leg.  Kind of a funny definition, but I am sure they document it someplace.

Overall, I think I will stick with Real Simple for telling me how best to prepare kale, and leave the traveling salesman information to other sources.

[Thanks to Ilona for pointing out the blurb in the magazine.]

(*)  Our favorite recipe so far has been “Scallops with Sweet Cucumber and Mango Salsa”.  Call it the official recipe of Michael Trick’s Operations Research Blog!

Serves 4 Hands-On Time: 25m Total Time: 25m

Ingredients

  • 1 cup long-grain white rice (such as jasmine)
  • 2 mangoes, cut into 1/2-inch pieces
  • 2 Kirby cucumbers or 1 regular cucumber, peeled and cut into 1/2-inch pieces
  • 1 tablespoon grated ginger
  • 2 teaspoons fresh lime juice
  • 2 tablespoons extra-virgin olive oil
  • 1/2 cup fresh cilantro, chopped
  • kosher salt and pepper
  • 1 1/2 pounds large sea scallops

Directions

  1. Cook the rice according to the package directions.
  2. Meanwhile, in a medium bowl, combine the mangoes, cucumbers, ginger, lime juice, 1 tablespoon of the oil, the cilantro, 1/2 teaspoon salt, and 1/8 teaspoon pepper; set aside.
  3. Rinse the scallops and pat them dry with paper towels. Season with 1/4 teaspoon salt and 1/8 teaspoon pepper. Heat the remaining oil in a large skillet over medium-high heat. Add the scallops and cook until golden brown and the same color throughout, about 2 minutes per side. Divide among individual plates and serve atop the rice with the salsa.

Open source replacement for Solver in Excel

Andrew Mason, who I got to know very well during my year in Auckland, has put together an open source Excel add-in that extends Excel’s built-in Solver (a product from Frontline Systems) and replaces the underlying optimization engine with the optimization code CBC from COIN-OR.  OpenSolver allows the solving of models without limits to the number of variables or constraints.  Andrew has also provided some enhancements including a model visualizer and the capability for fast reoptimization.

OpenSolver works with linear and (mixed-)integer programming models, but not with nonlinear models.  So keep those @IF and @POWER functions away from it!

This add-in a great illustration of both the strengths and weaknesses of open source.  On the plus side, all the work that people put into CBC can now find a much wider audience through the efforts of someone else.  No coordination was needed:  people like Andrew are able to use their own creativity and drive to put together something useful.

However, it is important to note that there are different open source licenses, and that Andrew has licensed this under GPL, while most of COIN-OR is licensed under CPL/EPL.  For most, this licensing difference is not of importance, but any work that is derivative of Andrew’s work can only be distributed under GPL.  Trying to figure out what is a derivative work is beyond me, but is important to those who would like to further develop systems based on Andrew’s work.  I note this fully respecting Andrew’s right to pick whatever license works best for him!

For me, this will be great to point to in class:  MBA student’s love Solver, but get frustrated with size limits.  I can now point to an easy to use add-in that removes those limits without requiring any changes in models.