I am trying to access as many resources as possible on internet, my universities databases/archives and now on to this wonderful forum. My primary reason to put this question is that I am looking forward to be initiated in the recent trends and scope of this topic. I am considering taking this topic as my research(which I have not yet decided upon). I would appreciate if the readers will post something on any of these: Books(Introductory); Industry needs and future role; Examples of problems; Success stories and Case studies; Guidance related to skills needed, various OR subjects etc. And please don't take it personally if it offends you in any way! Thanks!

asked 30 May '13, 14:32

phDcandidate's gravatar image

phDcandidate
102
accept rate: 0%

That's very broad questions, I think you will find several old discussions covering at least some of them. Start by searching old questions or try to be more specific.

(30 May '13, 14:36) Bo Jensen ♦

Can you suggest the keywords which might help me on this forum? I can't get any result just by typing "Large Scale Optimization"

(30 May '13, 14:39) phDcandidate

Depends on what you exactly mean by Large Scale, I mean is it just in the traditional sense that we are talking industrial sized problems or huge problem sizes ? For books, try searching on "Books". For skills try "Skills" (sorry I am not trying to be sarcastic here :-) ) Lots of good advices in the history, though I should warn you to listen to mine..

(30 May '13, 14:47) Bo Jensen ♦

This book on Amazon talks about real-time optimization and large scale nature of data(maybe not computation intensive): http://www.amazon.com/Online-Optimization-Large-Scale-Systems/dp/3540424598 I may be wrong but, to me I think this topic relates with the way we are moving towards "Big Data"!

(30 May '13, 14:57) phDcandidate

But then again... this book is from 2001 – and let me point you to the Turing's Invisible Hand blog (by Noam Nisan, Tim Roughgarden et al.), quoting an analysis (by Martin Grötschel, one of the book's editors) of performance improvements on "big problems" over ~2 decades.

(30 May '13, 15:36) fbahr ♦

What is your major? What kinds of large-scale problems are you looking at? Is it large-scale in terms of data, or in terms of equations/constraints? I work in the area of large-scale optimization, but I realize the term itself is rather vague. You may need to provide more information for clarification.

(30 May '13, 16:55) Gilead ♦
1

When you ask an open-ended question like this, especially as a new user, most people will ignore you unless you include what you have already found in the body of the question. If you do that, then you or someone else can make it into a wiki question.

(30 May '13, 23:07) Leo

I have started PhD in Industrial Engg (O.R.) having a bachelor in Mechanical Engineering. I am looking forward to start my Dissertation topic asap. And most important thing I want for that is that the research that I am doing is on the rise in industry. I am sure there are many more trending topics that one can pursue as of now, and I don't mind if you all want to share something on that too. But for the sake of this thread, let's only talk about the title. Think of what one can need to start their PhD in this field, how and where to start from, and advise me then, because that is my situation.

(31 May '13, 08:58) phDcandidate
showing 5 of 8 show 3 more comments

What is large scale optimization becomes "business as usual" 10 years later. I'd look for problems that are out of reach of commercial software solvers today. I see three main areas:

  • Real time optimization (also called online optimization) that involves midsize problems, ie problems that can be solved in minutes when the need is to solve them in milliseconds
  • Optimization with uncertainty. This includes robust optimization, stochastic optimization, etc. Just look for large classical optimization problems (that are solvable in, say, hours, with commercial solvers), and look for the uncertain version of these problems
  • Non linear global optimization. There, large scale starts with relatively small problem sizes.
link

answered 31 May '13, 03:43

jfpuget's gravatar image

jfpuget
2.5k310
accept rate: 8%

edited 01 Jun '13, 05:43

I was definitely fascinated by the concept of Real-time optimization. In addition to that, I am only interested in reading more about optimization involving lots of input data, not necessarily involving tremendous processing power(comment below by Nathan is important in this regard). Please shadow more light on this, and whatever resources come to your mind to this direction, Thanks!

(31 May '13, 08:50) phDcandidate

I am not sure I understand the reason to focus on "lots of input data", I realize my most hated buzzword "bigdata" has some broad attention, but I think @jpuget has a valuable point in "What is large scale optimization becomes "business as usual" 10 years later.". The three points mentioned are from a practical view very interesting, especially real time optimization.

(31 May '13, 17:39) Bo Jensen ♦

Two (still broad) interpretations of "large scale" are

  1. Lots of input data
  2. Takes forever to compute an answer

If you mean #1 then you will by definition be led to approaches that break problems into pieces that can be handled separately (and then reintegrated to form a full solution). You can find examples of such approaches by searching for "map reduce" or "big data machine learning". Some papers labeled machine learning are optimization papers in disguise. Most enterprises in most verticals are already trying to solve (or will soon start to solve) such problems. There was an article about GE and big data analytics this week.

If you mean #2 then as JF said above there a number of other subpaths you may pursue. Many (if not most) real world optimization problems can be made arbitrarily tough to solve by shoveling in more input data or increasing the realism (and therefore complexity) of the model. If you want to solve optimization problems faster you have essentially four choices:

  1. Bring more processing power to bear on the problem, (e.g. parallelize the math)
  2. Solve a simpler problem, (e.g. relax important restrictions)
  3. Tune your solution approach, (e.g. invent a new MIP cut)
  4. Invent a new solution approach. (good luck, and send me your code)

Nate

link

answered 31 May '13, 08:41

Nathan%20Brixius's gravatar image

Nathan Brixius
2122
accept rate: 0%

Your answer
toggle preview

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here

By RSS:

Answers

Answers and Comments

Markdown Basics

  • *italic* or _italic_
  • **bold** or __bold__
  • link:[text](http://url.com/ "Title")
  • image?![alt text](/path/img.jpg "Title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported

Tags:

×190
×1
×1

Asked: 30 May '13, 14:32

Seen: 2,173 times

Last updated: 01 Jun '13, 05:43

OR-Exchange! Your site for questions, answers, and announcements about operations research.