User samik r. - OR-Exchange

Answer by Samik R. for Where did OR people go ?

2010-07-23T19:56:50Z

Did you guys know about the "INFORMS Member Forum"? I just found out about that from the INFORMS newsletter. Looks like it is set up for similar type of questions.

We should seriously, have a discussion with INFORMS on this initiative and other similar and competing initiatives. Let's put the whole community's weight behind any one of the Web2.0 type initiatives rather than fragment it.

Answer by Samik R. for What to do when line search fails in Newton's method with approximated Jacobian matrix

2010-07-18T15:03:06Z

In my experience working with nonlinear optimization problems (not nonlinear system of equations), I have seen that direct implementation of Wolfe-Powell conditions are better than backtracking line search algorithms for finding step length. For one such implementation, see Section 3.5 of Numerical Optimization, 2nd Ed., Nocedal and Wright. This algorithm is actually and adaptation of the same from "Practical Methods of Optimization, 2nd ed." by Fletcher. I suggest you take a look at both the books and the line search paper by More and Thuente. My implementation was an adaptation of the Nocedal-Wright algorithm, augmented by a few things from the Fletcher book, and then I added some boundary conditions from the More-Thuente paper. You can also look at the code for More-Thuente algorithm, it is available inside quite a few open-source packages. Hope this helps.

Answer by Samik R. for Where did OR people go ?

2010-07-18T14:13:27Z

I love this site, and I use sci.op-research as well. With all the enthusiasm to promote and advance this site, I would like to point out that this exchange is still lacking hard-core OR discussions. I just saw one posted (What to do when line search fails in Newton’s method with approximated Jacobian matrix), and there are a few more, but if you look at the question set in the exchange, right now it is rather heavy on "Misc OR related questions". I will agree that "hard-core OR question" and "misc OR related question" are subjective terms, but I guess the sample I mentioned is the type of question I am looking for. In my opinion, sci.op-research still sees quite a few of these questions, when it doesn't see spam. We need to do something to redirect these questions to this exchange.

Question to Mike: am I correct that these type of questions are also the intended questions for this exchange?

Lastly, my thoughts on bringing the OR people back to this exchange (and also maybe get new members):
- We should create a flier for this exchange and post it in IEOR dept. boards. I have seen that this method still works pretty well in attracting attention to lot of people, who would probably go and visit once to find out what is the buzz about, and then, if they find out that it serves their purpose, may start contributing.
- We can have a talk in INFORMS annual, where we can share the statistics from the group, its objectives and how its improving. I have seen that this type of talks attract interested folks.
- Those of us in the industry (and also academia) can promote the group using a line in the signature.

My $0.02 (or may be slightly more).

Answer by Samik R. for What is your favorite Operations Research book?

2010-06-25T16:30:32Z

In my undergrad, I was taught using "Operations Research: An Introduction" by Hamdy Taha. Although I quickly outgrew that book, but I still think it is a good book for beginners. I understand that OP asks for favorite OR book, and I don't think I will consider this as that, but it is a good book nevertheless. And, from favorite OR book standpoint, unfortunately, I don't have any at this stage. Basically, I end up using 3-4 books per topic area, depending on which topic I am working on at any given time.

Answer by Samik R. for GPU and MIP solvers

2010-05-25T15:14:03Z

I attended a related talk in INFORMS09, a work by a student of Prof. Wright in Wisconsin. Check out this work from Sangkyun Lee. Here is the abstract:

Several highly effective algorithms for compressed sensing and image reconstruction applications can be efficiently implemented on modern graphical processing units (GPUs). The properties of the algorithms and applications are discussed, and computational results are presented to show large speedups over CPU implementations.

Answer by Samik R. for Eyjafjallajökull and OR

2010-04-19T20:52:46Z

I am sure the OR group in each of the affected airlines are busy right now rerunning their flight and crew scheduling models (since quite a few airports are going to open today).

Answer by Samik R. for CP and Integer programming

2010-01-06T05:53:01Z

There is an introductory paper available at the Interfaces, titled "Program Does Not Equal Program: Constraint Programming and Its Relationship to Mathematical Programming". A PDF link is here. May not be exactly what you are looking for, but I found the article very nice to start CP with.

Answer by Samik R. for RSS feed for this site

2010-01-06T05:48:41Z

What is the RSS feed for the site? I couldn't find it from the FAQ.

Comment by Samik R.

2010-05-25T15:03:19Z

The software is interesting, but not well supported. I once had a question about some stuff and never heard back.