INFORMS Annual Conference  

INFORMS Annual Meeting 2012, Phoenix, AZ, October 14-17

 
Blog:
blog home Home
 
       Follow:
RSS
 

Stochastic Thoughts from Sunday

by Paul Rubin on October 15th, 2012

Sometimes the real take-aways from presentations are not the details of the talk, but some remark a presenter made, or a question someone raised … or maybe just observations of the audience.

  • In the SA12 session (integer programming), James Ostrowski opined that antisymmetry constraints might make it harder to find a feasible solution by killing off the “clones” (my term) of any solution, leaving fewer feasible solutions to trip over. The same argument could be made for optimal solutions being harder to find. On the other hand, antisymmetry constraints may reduce dithering high in the search tree. So, on net, are they good or bad?
  • The SB08 session was the first of two sessions with presentations by the finalists in the “Doing Good with Good OR” competition. All three presentations were excellent (not universally true of sessions featuring more seasoned presenters), and the work seemed to be truly useful. Sadly, a schedule conflict ruled out my seeing the other session.
  • Two of the presentations in SB08 dealt with the systems in which a diagnostician interprets a test (or tests), leading to an estimate of the severity of an illness, which in turn triggers decisions about further diagnosis, treatment or whatever. Neither mentioned (at least that I heard) the possibility of using a diagnostician’s history of diagnoses and actual outcomes to modify the conversion of their assessment into probabilities for disease severity or progression. In other words, if diagnostician A is a bit of an alarmist and diagnostician B is possible a bit too laid back, that might be captured and used in a corrective manner. After all, it’s the age of informatics, right? (At least until next year’s meeting.)
  • The third presenter mentioned the difficulty in getting accurate census/demographic data in some poorer countries. I recall reading somewhere that there is a surprisingly high (to me) rate of ownership and use of cell phones in many impoverished areas. Cellular service providers might be able to supply data on the number of distinct phones used in different parts of the country (the cell towers provide locations). If available, could that data be used to sharpen estimates of populations by district or region?
  • I found it a bit odd that I personally supplied nearly all the gray hair in the audience. These talks were both inspirational and examples of how students can be engaged in important OR projects. That engagement will typically need support and encouragement from faculty. Are junior faculty more receptive to innovations like this than senior faculty?
  • SC16 was a tutorial on OR in humanitarian logistics. I attended a similar talk last year, but the content of this year’s talk was fresh (to me). The session was very informative and, for me, enjoyable. The audience hovered around 40, which is decent but less than my prior expectation given the social significance of the work. This may mean that previous efforts to get the word out have already exposed a large portion of the membership to it, or it may mean that future sessions need to be advertised more aggressively. It could of course just be a coincidence. (Was somebody giving out free food in another session? Tell me I didn’t miss free food!) It may also be that the scope of engagement in humanitarian logistics problems, and the need to work with actual end users, turn people off to the topic. I hope not, although I understand the concern about how much effort it would take to get started.
  • One of the issues that came up in SC16 was the need to get from models suitable for academic papers to simple but functional decision support tools (software) that NGOs can use to put those models to work. The academic system tends to skew rewards toward publishing papers and not toward implementation. (As a sidebar, I’m involved with the COIN-OR foundation, which deals with open source projects. We see a similar pattern there. Project leaders can be rewarded for writing and publishing code. Documenting and maintaining it is neither fun nor intellectually stimulating nor, typically, rewarded.) Getting back on topic, I wonder if there are opportunities to pair OR people (faculty, maybe retirees) with non-OR socially conscious programmers, the former explaining how the computations need to be done and the latter hacking code to implement the models?

Comments are closed for this entry.