3 ms·

Care to share your dissertation? If I'm guessing correctly what scenario aggregation is by just the name, I've found that technique is very powerful when doing

Care to share your dissertation? If I'm guessing correctly what scenario aggregation is by just the name, I've found that technique is very powerful when doing that with plain old human-based decision making. Basically if you have a bunch of complex scenarios that you're trying to consider at once it can be very difficult to make progress because you have to compare whole trees of options against each other along with their probability every time you make a change anywhere. But if several scenarios have some prefix where they are basically equivalent (or can be made to be), you can bundle them together and consider the group as a whole, putting off deciding between them until later. Sometimes this can collapse a huge complex decision tree into the one obviously correct thing to do right now, until the point where they diverge again.

graycat 5y ago
Good, fast view of scenario aggregation. The idea is powerful, not just a heuristic, and from R. T. Rockafellar. I didn't want to publish my dissertation because that would be like giving it away for free. Instead I wanted to sell it! I made some efforts to sell it but ran into the problem I mentioned for the suits. So, it didn't sell. But I still don't want to give it away for free.
- infogulch 5y ago
  I've never heard of selling a dissertation, but this seems to be a reasonable case for it. I guess you're selling at a fairly high price otherwise it would just be a book, no? I can imagine one fear such suits might have is that they would invest in access to this idea but be unable to operationalize it. Perhaps a startup to sell good decisions (a funny thing to type) is in order. Hit me up on the email on my about page if you'd like to chat more. :) Good luck!
  - graycat 5y ago
    Thanks for the offer. Before my Ph.D., I had a good career going, non-academic. At times, math was a help, so I got a pure/applied math Ph.D. But I never had any intention of being a professor; I got the Ph.D. to help the good career I already had. Yup, mostly Ph.D.s are guided to think that their future is as a prof and there, with "publish or perish" they should publish everything they can. Fine for them. But for me, did I mention, I had no intention of being a prof. For selling the work, the work was to help some big companies in a big industry save some millions of dollars a year. For me, I assumed that the world class university where I got my Ph.D. would provide significant credibility and that the chance of such big bucks savings would at least get me a lot of conversations. Nope: I got a lot of silence and otherwise some resentment, "We are already doing the most best possible optimal thank you very much." or some such. Net, the selling I tried didn't work. I didn't have the money, selling skills, or insight into organizational behavior to continue trying to sell. If they regarded my world class research university Ph.D. and dissertation as junk to be tossed in the trash, with no investigation at all, then they were too tough for me to work with. Thanks for the offer, but I gave my view of applying Markov decision theory -- nearly always in practice, it's usually too difficult technically and otherwise still not wanted or even deeply resented. So, the whole subject is just a waste? Maybe. Here is a mature view: The best of research is about the best there is in civilization. Still, nearly always the world of practice would like to see research just go away. For Markov decision theory, a reasonable research question was, what can be done? Can we find optimal solutions in any sense? If so, can those solutions make money in practice? So, for the good news, the research was successful: Yes, we can write out in great detail, highly polished, how to get optimal solutions. So, we wanted optimal solutions, and now we know what optimal solutions look like. Or, now the research is done. And, now we have removed all doubt: Essentially always, optimal solutions are too challenging technically. So, we can know we should find something else to do. Now for the bad news: Now that we do know what the optimal solutions look like, we also know that the solutions are too challenging technically to be applicable in practice except for some rare situations. So, for 99 44/100% of practice, Markov decision theory and a dime will just cover a 10 cent cup of coffee. Thanks for the offer, but I'm doing a startup, one that has essentially nothing to do with Markov decision theory.

graycat 5y ago

Good, fast view of scenario aggregation. The idea is powerful, not just a heuristic, and from R. T. Rockafellar. I didn't want to publish my dissertation because that would be like giving it away for free. Instead I wanted to sell it! I made some efforts to sell it but ran into the problem I mentioned for the suits. So, it didn't sell. But I still don't want to give it away for free.

infogulch 5y ago
I've never heard of selling a dissertation, but this seems to be a reasonable case for it. I guess you're selling at a fairly high price otherwise it would just be a book, no? I can imagine one fear such suits might have is that they would invest in access to this idea but be unable to operationalize it. Perhaps a startup to sell good decisions (a funny thing to type) is in order. Hit me up on the email on my about page if you'd like to chat more. :) Good luck!
- graycat 5y ago
  Thanks for the offer. Before my Ph.D., I had a good career going, non-academic. At times, math was a help, so I got a pure/applied math Ph.D. But I never had any intention of being a professor; I got the Ph.D. to help the good career I already had. Yup, mostly Ph.D.s are guided to think that their future is as a prof and there, with "publish or perish" they should publish everything they can. Fine for them. But for me, did I mention, I had no intention of being a prof. For selling the work, the work was to help some big companies in a big industry save some millions of dollars a year. For me, I assumed that the world class university where I got my Ph.D. would provide significant credibility and that the chance of such big bucks savings would at least get me a lot of conversations. Nope: I got a lot of silence and otherwise some resentment, "We are already doing the most best possible optimal thank you very much." or some such. Net, the selling I tried didn't work. I didn't have the money, selling skills, or insight into organizational behavior to continue trying to sell. If they regarded my world class research university Ph.D. and dissertation as junk to be tossed in the trash, with no investigation at all, then they were too tough for me to work with. Thanks for the offer, but I gave my view of applying Markov decision theory -- nearly always in practice, it's usually too difficult technically and otherwise still not wanted or even deeply resented. So, the whole subject is just a waste? Maybe. Here is a mature view: The best of research is about the best there is in civilization. Still, nearly always the world of practice would like to see research just go away. For Markov decision theory, a reasonable research question was, what can be done? Can we find optimal solutions in any sense? If so, can those solutions make money in practice? So, for the good news, the research was successful: Yes, we can write out in great detail, highly polished, how to get optimal solutions. So, we wanted optimal solutions, and now we know what optimal solutions look like. Or, now the research is done. And, now we have removed all doubt: Essentially always, optimal solutions are too challenging technically. So, we can know we should find something else to do. Now for the bad news: Now that we do know what the optimal solutions look like, we also know that the solutions are too challenging technically to be applicable in practice except for some rare situations. So, for 99 44/100% of practice, Markov decision theory and a dime will just cover a 10 cent cup of coffee. Thanks for the offer, but I'm doing a startup, one that has essentially nothing to do with Markov decision theory.