Desi_Enigma

Can you specify which program you are applying to at these universities? I'm in an EE PhD program working on statistical signal processing at the moment, but my interests have shifted more towards discrete optimization. I'd still like to be involved with some ML though, so shifting to OR/Stats seems ideal for me. I'm also looking at some of the programs you mentioned - particularly Cornell and Columbia. Are you aware of any other OR program which has lots of emphasis on both optimization and applied ML. I'm finding it hard to identify programs good at both - e.g. GATech is good at the former but weak in the latter. Duke seems strong in ML but doesn't have a strong optimization group. Let me know your thoughts

Just a few additional points, so that my previous post can be understood in context. My points were primarily targeted at this line. You should realize that this is simply too broad, and there are tons of areas (the ones I mentioned) which look at subsets of this problem. It is also not humanly possible for someone to address all the different problems which can be categorized according to your statement. For example, I am an EE major who focused primarily on statistical signal processing. These ideas are used in the context of EE to do channel estimation, stochastic control, synthesis of adaptive systems etc. all of which are great examples of "real-world" or practical problems. Each of these are used in day to day life ranging from communication, guidance and navigation, to industrial automation. People use exactly the same mathematical tools in departments like OR, statistics etc. for financial time series analysis. In applied math, some version of this could be used for weather forecasting, modelling complex systems (e.g. systems biology) and a whole lot more. All of these are using statistical signal processing to solve "real-world" problems. But the problems addressed are vastly different though mathematically, they are extremely close. So the real question is which real world problem are you talking about? I suspect that you have just come out of the shell of a pure math major, and are trying to characterize problems and programs as "pure or traditional" vs "real world or non-traditional". This cannot be further from the truth. The number of fields where people use math to solve practical problems far out number pure math or pure science programs. "In other words, I am more interested in the process of solving the puzzle than the pieces themselves." - this is not new at all, and comes across as very broad and ill-informed. I would disagree on this point. Though the core math tools used are nearly the same, it is not possible for someone trained only in the math tools to actually solve real problems. A key component which mathematicians overlook is the domain knowledge. Though the math tools would allow one to move across different fields, it is unlikely that they would know which problems are worth solving, which approach or solution is acceptable or practical, and which ones are nonsensical. For this reason, if you just have the tools, you can never "own" your project, and would likely be "contributing" to someone's project. You can keep moving from one project to another, helping out with your tools and ideas, but never owning any project or actually identifying a practical problem to be solved. If this is what you want to do, an applied math degree will fit your bill perfectly. On the other hand, if you have some burning question, and want to solve a real world problem, it is important to develop expertise in the math tools as well as develop domain knowledge. (A vague example would be something like solving energy crisis, how one can use mathematical methods for scheduling to minimize wastage, load balancing etc.) I have seen students with interests spanning across the entire spectrum, so your first job should be to figure out where you fit in. Once this is done, the choice of program will come out by itself. I'll also add a comment that there is at least one acceptable department or program for each point in this whole spectrum.

Hi Transformiao, I don't mean to hijack this thread, but I have created a similar one to discuss about strengths and weaknesses of top OR PhD programs. Since you did apply and got into some of the top OR programs (Columbia & Cornell), I was wondering if you could add your thoughts on the same. Others interested in statistics/OR are also most welcomed to comment. Best wishes for a great grad school experience, Enigma

Hi MarcusSolarz, I can totally understand and relate to your motivations and expectations. I came to the US two years ago for a PhD in EE at USC. I would soon be transferring/shifting to a PhD in applied math or operations research. I have some suggestions which might be useful for you. First, I should I say largely agree with many points raised by TakeruK and Igotnothin. As TakeruK pointed out, the idea of doing PhD is to become a world leader in research. There is no other reason to do a PhD. However, research is by no means is confined to academic research, and does not necessarily have to culminate with a publication. For example, quantitative finance in wall street firms have very high research activity. However, you'll never see a paper coming out of there (possibly a few patents). A similar argument can be made for some industrial research labs like IBM too where the number of publications do not reflect the research activity. Research and innovation happen in a variety of ways, and PhDs are by no means confined to academic research. I also agree with Igotnothin. Biostats/stats is a very active field solving "real-world" problems. Naturally, most PhDs in this area end up at industrial labs solving very "practical" problems ranging from pharma to experiment design for clinical trials. However, I would offer a word of caution. There are many fields that specialize in applying math to solve real problems. This is by no means a new thing, and such programs, thoughts, and ideas have existed for decades! Some such programs are applied mathematics (duh!), operations research (for some reason, not many are aware of this wonderful field), statistics (the flavor varies from place to place), and virtually all branches of engineering. All these vary in flavor and what sort of real world problems they want to tackle. I suggest that you spend time in realizing which real world problems you want to solve, because there are too many problems in the world, and each requires a different skill set. For example, I am switching from engineering to applied math, which should tell you that the nature of problems tackled are quite different, though most of these areas largely use the same or similar mathematical tools. Applied Math (aka computational science) basically develops math tools to help people with "domain expertise" (generally engineers) solve problems. This is ideal for someone who wants to build techniques and be involved in solving real world problems, but don't want to get their hands dirty with the actual problems. This might fit your requirements very well since a) you majored in math; b ) it teaches skills you want to learn (computational modeling, data science, fast solvers for large problems etc.); c) Is aimed precisely at cultivating interdisciplinary ideas to help solve real world problems which people with traditional training are not equipped to by themselves. Operations Research (OR) uses techniques from optimization, statistics, and engineering to solve a wide range of practical problems. The problems addressed in this field tend to have a "business" flavor, and a good chunk of PhDs end up in quant finance or related areas. It asks questions like how should I design my supply chain, how should I schedule jobs to prevent bottle necks, optimal resource allocation problems, decision making in face of uncertainty etc. Optimization and stochastic modelling (probability theory) are the "core" of OR. Statistics largely works with problems involving data. Some of these problems are also addressed in OR, and for this reason OR and Statistics are mostly combined into one unit, or work very close to one another. Pick the sort of real world problems that you like the most, and see where they best fit in the spectrum. There is of course considerable overlap between the fields above (in tools used), and people do move between them after PhDs. I know a few OR and biostat PhDs working at Facebook on social network analysis, many EE PhDs in quant finance, and quite a few applied math PhDs in drug design. I think you can develop a good set of skills through coursework and actually solve real world problems in your PhD itself if you work in above areas (I can personally attest to OR). Enigma

I am planning on applying to a few OR PhD programs this Fall and though about compiling a list of top IE/OR programs by major research fields. I figured the best way will be to post it here so that others can share their thoughts, and it will also be useful for students applying in the future. A quick background, I am going to start my 2nd year in a Southern Cal. EE PhD program. My interests have more or less shifted towards optimization, stochastic processes, and simulation. Also, given my performance over the last year, I think I can move to a better university. So figured I might as well apply, and if I catch a big fish, can switch next year picking up an MS en route. Research Field: Optimization, Stochastic Processes Tier 1: MIT, Stanford Tier 2: Columbia, Princeton, Cornell Tier 3: UC Berkeley, Georgia Tech, U Michigan, Northwestern, UIUC Tier 4: USC, Purdue, Wisconsin, Penn State ... Some stray observations - UC Berkeley seems to have a very small department with only 15 odd faculty. Most of them seem to be old and close to retirement, and hence there are only 6-7 "active faculty" each working on almost orthogonal topics. Not sure why it is ranked so high, am I missing something? Also, GATech seems to have the same problem too. Most of their star optimization guys are 65+ and might retire anytime. Taking those 2-3 stars away, GATech's optimization and stochastics groups don't seem that impressive (compared to what they claim, #1 program). Cornell and Princeton seem to be surprise packages - I didn't expect them to be this strong. It would be great if other students can share their thoughts, and rank-tier the programs based on their research field (financial engineering, production, logistics, healthcare etc.) It would be great if some Fall 2015 applicant can also share their thoughts based on visit weekend, interviews etc.

I'm here at LA, so I guess he has some preference to UCLA, USC, and UCSD. From what I gather (I haven't personally looked into this), UCSB is not that strong in machine learning as far as CS goes. But I heard they have a kick-ass signal processing group in ECE, so he "might" be interested, though a question of research fit might crop up. Bay area is also a fair game since we have some working class family there. I think he will shoot an app to UCB too, but I am not too optimistic about those chances. I am also kind of talking him into applying for UT Austin. I think he has a fair chance there, and I have heard that the ML faculty is strong. Also, I have heard from many friends that Austin is very Indian-like (climate, culture, demographics etc.) so he might be OK there. The concept of education loans are not well established in India. Even one semester unfunded is kind of a big deal when you factor in the exchange rate. Also, loans in India are against a collateral. So we basically need to take a loan against some asset, which in our case is just a house. My parents would rather have him do his PhD in India itself instead of spending a hefty amount when he is not particularly fond of coming here. Funded offer is a different ball game, the worst case scenario then is just one wasted year. If he is that bad in adapting to a new place, he can simply go back and do a PhD in India. As I said, he is kind of an exception - an extreme case of introvert, hyper, and shut type of person. Thanks for your inputs

This is a classic. He brings up reason after reason which don't really make any sense. But, I can clearly see where his problem lies. He is just scared of moving away from home, and he won't admit it. Some people may not be able to relate to this, but lifestyle is completely different in India. Parents take care of everything for kids; there are servant maids for everything; people have drivers etc. Some spoilt kids find it very hard to adapt (I personally found the first couple of months quite hard ) Most Indian students have this problem, but my brother is kind of an extreme case. By nature he is quite scared and an introvert, right from childhood. So, I don't really know how to convince him. He will also never admit that this is his problem, but would rather point out some drawbacks of all other universities (while these five would have pretty much the same issues). No, he is pretty set on PhD. Also, we can't really afford an MS. I barely scrape through with my stipend, and my parents have no intention of taking out a loan. Yes, LORs will be a problem. Professors from different departments discussing the application was exactly the point of the question. Does that happen? Are you saying that, if he is rejected by one, his chances of getting into another department is slim (or lesser compared to an application at another school of similar repute)? Regarding his profile, I think its reasonably strong. Stanford and CMU might be out of his league (MS possible, but we can't afford) but I guess the rest are fair game. But you never know with PhD applications :\

Hello friends, Its my younger brother's turn to apply for graduate schools, and he has some questions for which I don't have comprehensive answers. So, I thought I'll post it here and get your opinion. Short question: Are the odds of admission success when applying to multiple programs in the same university, and applying to different universities, the same or comparable? (i.e. chances of admit when comparing application to programs A & B in school X; vs applying to one program in school X and one program in school Y - assume X and Y are of similar ranks/reputation) He is interested in doing a PhD related to optimization, signal processing, and machine learning. The thing is, he has limited his options based on location. Basically he wants to study either in California (I'm here and Cal has same weather as back home in India) or at CMU (have a cousin there). Basically, he is a wuss and is kind of scared about moving to the US. He is willing to come only if he can be close to friends/family, which is possible only in California or CMU. He did his BS in ECE, and considering his interests, he doesn't mind a PhD in any of EE, CS, Statistics, or Operations Research. Is there any use in applying to more than one program at the same university? Say he applies to both ECE and CS at UCSD, are the applications considered "separate", so that he won't get outright rejected for all the programs. CMU is also similar where one can apply for multiple programs in the school of computing. Is there any use in applying to more than one program in the same university or is it just a waster of money? Basically he just has 5 universities in mind (Stanford, CMU, UCSD, UCLA, USC) and wants to get into one of these. He doesn't mind the program as long as it is one of the above mentioned. Normally people suggest applying to at least 8 programs to be safe. Does applying to multiple programs (thus bringing up the total number of applications to 10) increase the odds? Is he putting all his eggs in very few baskets? Please share your thoughts.