“[I]t seems to me that the term `linear programming’ is a most unfortunate phrase for this promising technique, particularly since many possible extensions appear to be in nonlinear directions. A more general yet more descriptive term, such as `bounded optimization,’ might have been a happier choice.Philip M. Morse (1953). Trends in operations research. Journal of the Operations Research Society of America, 1(4), 159-165.
It’s interesting that the term “linear programming” has been disliked since it was introduced.
It’s also interesting to hear skepticism about the usefulness of linear programming. This reminds me of a story about George Dantzig in 1948, who also met skepticism regarding the usefulness of linear programming. Read my post “Happiness is assuming the world is linear” for more information.
What do you think of the term “bounded optimization?” Is it better or worse than the term “linear programming?”
May 16th, 2022 at 10:10 am
I think bounded is a very unfortunate word. It doesn’t really capture the inherent differences in solvability. Not surprising though, without actual computer implementations it was hard to grasp what could be solved and how fast. In the absence of that it is hard to see that linear/quadratic/conic/convex/etc. is the important feature. (Just a speculation, but I think Morse was referring to the fact that the optimal solution is obtained as the intersection of some of the constraints vs inside the feasible set, where it could be located with calculus.)
I guess these days it would be called constrained optimization, which is an equally bad term. All (practical) optimization problems are constrained, even the seemingly unconstrained ones.
May 16th, 2022 at 10:32 am
My reason for disliking the term is that the distinction between linear and nonlinear optimization is much less important than the distinction between convex and nonconvex optimization.
May 16th, 2022 at 12:25 pm
I like the fact that almost all fundamental optimization techniques and algorithms have unassuming names (maybe with the exception of “dynamic programming”, which does not have a modest origin, although it is not arrogant either). In contrast to some disciplines that enjoy giving their methods important names, I find it interesting that optimization has made its way until the very core of our society without “naming paraphernalia”. I like the attitude of making silent progress, in contrast with the overpromising and the marketing efforts (followed by disappointment and lack or interest/funding) that are unfortunately sometimes seen in some disciplines.
May 16th, 2022 at 12:59 pm
“Bounded optimization” could mean almost anything (optimal control theory?). My concern with “linear programming” and, even more so, with “integer programming” has less to do with linearity being stifling and more to do with the ambiguity of “programming”. I’m pretty sure the term was chosen as a reference to “program of activities” well before the computer era, but today a lot of people hear “integer program” and think “computer program that messes around with integers” … which is not wholly inapt, but still rather misleading. If someone outside the OR community asks me about research interests, I fall back on “discrete optimization”.
May 19th, 2022 at 11:25 am
If I recall correctly, George Dantzig coined the term Linear Programming. Much of the confusion is due to the word ‘programming’, which now is more closely associated with computer programming. However, I believe Dantzig used the word programming akin to a recommended plan to follow, the way today we think of a program of study at school or a program for theater or a concert. Irv Lustig and Jean-François Puget wrote a paper on this naming issue in Interfaces: https://pubsonline.informs.org/doi/pdf/10.1287/inte.184.108.40.20647
Perhaps a better word could be “plan” — linear planning, integer planning, etc.