20 February 2009

Balintfy made the world better with OR.

Joseph Balintfy has died. He was a professor of Management Science and Operations Research at the University of Massachusetts, Amherst. What did Balintfy do that was so cool? Balintfy studied using O.R. to improve nutrition.

Back in 1945, George Stigler tried to calculate the minimum cost of subsistence. He found a good low-cost diet, but he could not prove that it was the cheapest, which also satisfied all nutritional requirements. A couple years later, George Dantzig set about solving this problem, inventing the simplex algorithm to so. This is all well-worn history to the legion of operations researchers.

What is also well known is that Dantzig's solution was awful: liver, cabbage, corn meal, evaporated milk, lard, peanut butter, potatoes, spinach, and wheat flour. And if we update all the nutritional information, Garille and Gass showed that it's not going to get much better.

Balintfy added palatability. Rather than finding the minimum cost of substistence, Balintfy studied the cost of decent subsistence, with considerable implications for national policy. Balintfy wasn't the first person to study this, as Benson at Rutgers was also researching this as early as 1960. (Balintfy even gave credit to Benson for this.) But Balintfy did a great job of it, studying computerized menu planning, food preference over time, and even subtle issues like the food mix for best bioavailability of nutrients, and the probability of satisfying nutritional constraints, given a random distribution of nutrients in a given food. He also studied food price indices, noting that existing fixed weight indices do not allow for substitutions as relative prices change.

Balintfy wrote a complete menu planning system called CAMP. With regal foresight and generosity, he put his program in the public domain. He saw that his program could help the world eat more nutritiously and more cheaply, so he gave it away! How cool is that! How cool is that! (In stark contrast to the breath-taking chutzpah and greed of Anjan Ghose.) According to Balintfy's 1975 article, the program has been used to feed over a million people, and that was 34 years ago. Because he put it in the public domain, others have picked it up and put it in a variety of commercial programs, and the number of people who have benefited is now countless.

However, I am not aware of it being used anywhere now. Also, I haven't been able to find Balintfy's code, but have started to ask colleagues about it. It might be a nice addition to the open source COIN-OR initiative, which was started by IBM; all the more elegant as Balintfy released his code through IBM's Contributed Program Library. It must have been one of the very earliest "open source" programs!

I've been studying problems of nutrition and diet for some years, and my nascent Wagmu site is part of the result. It doesn't have all the tricks that Balintfy implemented, and may never. I'm simply not as clever (nor as well funded) as Balintfy was.


Jo Jordan said...

Couldn't get in John.

Filled in lots of info. Got the email and link. Went back to a login screen - huh? what is the username - real, nickname( was there one), email.

Too much to remember!

John F. Raffensperger said...

Thanks, Jo! The sign in form lets you put in a username at the top. You can use that username to sign into Wagmu. Eventually, we'll change over to a email-is-username approach, but please don't wait for that! And thanks very much for your interest.

Maria Kirby said...

I have a small problem with these kind of nutritional calculations: they don't, maybe can't, take into account the variety that a hunter/gatherer might eat. As I see the long term sustainability issues, we need to move to an agricultural form that is much more like the early hunter/gatherers. We are at the top of the food chain. We are the ones that are smart enough to keep ecosystems in balance, taking only the excesses that nature provides. That means eating many different kinds of things. The domestication of a limited selection of herbaceous and animal food sources is a recipe for eventual extinction, both for the ecosystems that agriculture replaces and for the agriculture itself.

John F. Raffensperger said...

Hi, Maria,

Thanks for your comment! Variety is good for us, as it helps ensure we get all our nutrients. As for environmental impact, the situation is a little more complicated, but fortunately very positive.

When we run these minimum cost models, they tend to omit meat and dairy, which are the foods with the big environmental impacts. Meat and dairy hurt the environment more than vegetables by an order of magnitude. So if you gave up on a variety of yogurt, cheese, pork, and beef, and instead ate just green peas, you would help the environment tremendously!

In any case, Balintfy's model considers variety in clever way: palatability is a function of how recently a given food was eaten. So the solutions tended toward variety, too.

Anonymous said...

I found a reference to Joseph L. Balintfy during my search for binary programming algorithms. I needed an algorithm that works with multiple-choice classes. After reading an article "Analysis of algorithm for the Zero-One programming" I finnally succeeded to get the Balintfy's article "Menu planning by computer". Do you have any experience with this algorithm?
Roman ZOGATA, Czech Republic, roman.zogata@gmail.com

Joe Balintfy said...

Thanks for this post about my dad. I've linked to it from a new blog I started recently at www.balintfy.org.