We would like to thank everyone who has responded to our request for NEOS stories. Hopefully this page will provide an idea of the range of applications people are attacking with NEOS. If you would like to contribute to this page, send us a message describing the problems or problem types you've solved with NEOS and the solvers used or fill out the form you will find by clicking the Comments and Questions button at the bottom of the page. The comments here are roughly categorized into Education, Business & Finance, and Science & Industry sections.

Education

I have been using the mixed-integer solvers at NEOS to study the complexity of phase retrieval, a problem encountered in several disciplines, principally crystallography and astronomy. My NSF grant has a budget for computer software, but not in the amount required to purchase some of the high end optimization software at NEOS. Another benefit provided by NEOS is that I am able to do serious number crunching from my humble laptop.

NEOS has been a very valuable tool in the two graduate optimization courses that I regularly teach: CS 719 Network Flows and CS 720 Integer Programming. NEOS allows students to see a broader variety of solvers than we have available in the Computer Sciences Department at the University of Wisconsin - Madison , and also helps to overcome a variety of time/storage limits that apply to student jobs here. For example, the student version of AMPL packaged with the text for CS 720 allows at most 300 constraints and 300 variables, but several students in the course are doing course projects involving problems that are substantially larger than those limits.

I have been teaching a course on linear and integer programming over the past two months. Part of the course is based on projects where the students use GAMS. Previous years, they have used the student version of GAMS to model the project problems. This year, I have given them larger problems where they have sent their GAMS files to NEOS and used XPRESS-MP to solve the problems. The ability to use GAMS and not being limited in size by the student version has been extremely useful. I find NEOS a very useful tool for teaching this course. When the students leave the course, they know how to build a GAMS model, and they also know that they can solve large problems via NEOS.

Royal Institute of Technology, Stockholm

We very much appreciate the NEOS server service. We find that not only are the solvers and the software reliable, but also that the people maintaining and developing the codes are very helpful. The NEOS server provides the possibility to run problems without going through the whole process of installing a software package, paired with the possibility to run instances on powerful machines.

At the university of Cologne we are interested in finding good bounds on the maximum cut value in sparse graphs. This problem has an application in theoretical physics and is interesting to the physicists' community. We calculate an upper bound on the exact value of the maximum cut by formulating it as a positive semidefinite program. We then add valid inequalities to tighten the relaxation. We solve the resulting problem exactly with the solvers CSDP and SDPA.

I am a PhD student at the University of Melbourne, Australia. Recently, I have been using NEOS (solvers:PATH, DONLP2, FILTER, MINOS) to carry out some numerical tests on a game theory model arising in currently debated electricity restructuring.

A big thank you to NEOS staff for providing such a wonderful facility (free!).

Excellent tool! I am using it for academic research regarding route selection in packet switched networks. I had no prior experience in the gams language but was able to quickly hack together a series of equations that the nlp solver was able to give me good results with.

I would like to thank all the people who provide and maintain NEOS.� At the beginning of this summer I began a term of assisting a professor at my university with his research. The daunting task of solving nonlinear programming problems was put on my shoulders and I felt as though I was in over my head.� I didn't even know what nonlinear programming was and after I discovered what it was, it became clear how enormous a task it would be to solve the problems assigned to me.�I bought Maple 6 to try to help me, but the calculation times were�huge, as well as the round off error.�Then looking for a better way,�I found NEOS.� With the great abundance of help given me by the caretaker(s) of NEOS, it took a short while to figure out how to setup the input file (I used AMPL format).� I had extremely complicated objective functions, both convex and nonconvex,�an arm load of variables, and an armload of convex, nonconvex, equality and inequality constraints, but when I sent off the information via the web submission form, within seconds I received extremely accurate and consistent results.� The results were used for verifying a certain theory�in my professor's research and so accuracy was extremely important.��It has been an invaluable service to me and I truly doubt that I could have done my job to the satisfaction of my professor.� I would highly recommend NEOS to anyone needing constrained optimization calculations done quickly and accurately.

I am a graduate student at RPI in Troy, NY working on non-linear polymer diffusion in confined spaces. Specifically, I am attempting to find solutions to the modified Cahn-Hilliard equation, a fourth order partial differential equation by techniques to minimize the total free energy of the system. While these problems are certainly not as complex as the large scale process optimization studies performed by other groups, the NEOS system using the Lancelot solver has allowed me to spend more time on experimental work that is also of interest to us since the optimization routines are made easier to use than if I had to write them myself. The submission process is straightforward and the availability of the NEOS solver has allowed me to concentrate more on the physics of our problem than the mechanics of writing optimization codes. If programs like the IBM data visualization explorer and HDF were made as easy to use as the optimizers from ANL, I would be ecstatic. As it is, I would encourage the continued funding of the NEOS server at ANL and its related costs. The service has been a great help in my progress toward a doctorate.

This is a great resource for those of us who need to demonstrate large scale optimization to students, but who do not have enough need for this (or funds!) to justify the purchase of a large scale optimization package.

I have been using NEOS in my nonlinear programming course every spring semester. We use the student version of AMPL earlier in the semester, and then I assign a larger applied problem (that is too large to be handled by the student version), which the students are asked to model with AMPL and then solve. In the past I have used AMPL and MINOS, but in future semesters I plan to ask students to compare different solvers.
I have also used NEOS in assisting a non-optimization colleague of mine to solve optimization problems relating to database design.

Hi, I've been using your server to solve complementarity models of electricity markets (using the PATH solver and AMPL). This is academic research. Following is an abstract of my paper. Your site has been invaluable for me. Keep up the good work.

This paper presents a modeling framework for analyzing competition between multiple firms that each possess a mixture of hydroelectric and thermal generation resources. Based upon the Cournot concept of oligopoly competition, the model characterizes the Cournot equilibrium conditions of a multiperiod hydro-thermal scheduling problem. Using data from the western U.S. electricity market, this framework is implemented as a mixed linear complementarity model. The results show that some firms may find it profitable to allocate considerably more hydro production to off-peak periods relative to peak periods. This strategy is a marked contrast to the optimal hydro-schedules that would arise if no firms were acting strategically. These results highlight the need to explicitly consider profit maximizing behavior when examining the impact of regulatory and environmental policies on electricity market outcomes.

I had a large very non linear problem. I used NEOS because of the AMPL interface and the access to various solvers. Originally I had no luck because the problem was just too large. I condensed it to one with about 11000 variables and 11000 constraints, of which about 10000 variables and 3000 constraints were nonlinear. After considerable work on the parameters, MINOS solved many variations of the problem in about ten minutes elapsed time each. I'd like to thank the NEOS team for their prompt and helpful attention to some queries I had. It was great to have the opportunity to try my problems on a number of different solvers, although in the end, MINOS proved the most useful solver for this set. I also found the web interface and email results made it easy to run a large project remotely and the overall documentation is an excellent introduction for beginners to math programming.

I used the solver CSDP via NEOS to solve several dozen semidefinite optimization problems. These SDPs were very large, typically with over 100 variables and several thousands of constraints.

I found that the NEOS system was easy to use and quite reliable. I would certainly consider using it again in the future, as well as pointing it out to other people whose work also requires solving optimization problems.

Thank you again for setting up NEOS and for keeping it running!

We (my students and I) have been solving large semidefinite programs (SDPs) using both SeDumi and CSDP solvers. Both have worked excellently. We have been able to solve some very large problems. The problems arise from relaxations of quadratic assignment problems.

I just would like to thank you all for the neos server: it's just perfect. I learned AMPL to use the server and this was useful for its own sake. Now I use the server for teaching, letting the students to put hands on simple parametric optimizations. Moreover I use the server to solve problems for research purposes.

One of the main motivations was the ability to run very large (more than 1M arc) network flow problems. Student accounts here have rather limited memory allocations, and many students had problems just storing the input data files for 1M arc problems (and the storage for the temporary files used by the algorithms is several times larger than the input). I wanted students to develop an awareness of the size of the problems that network codes can solve, and this would have been hard to do locally.

Additionally, I wanted students to develop a feeling for the relative speed of various network and general LP codes. NEOS provides easy access to 2 different network algorithms (both of which I teach in my course) as well as LP codes. This aspect I could have handled locally (with problems much smaller than 1M arcs), but students are generally reluctant to learn multiple interfaces, so it was very convenient to do it all via NEOS.

I think that first we all should thank you and your colleagues for the excellent work done with NEOS. It is a great service to the public. I myself used it in trying out the AMPL interface for my code donlp2, most of the other codes being installed locally here. But I know of a colleague which uses NEOS as a teaching machinery for his nonlinear programming courses, which could otherwise be done only by paper and pencil due to the lack of appropriate computing resources.

I sincerely believe "optimization on the Internet", or some people refer to it in more general terms: "providing decision technologies over the Internet as services", is very important and will become ever more important in the future. For "optimization" to become a viable tool for the "end users", they must be provided with the latest optimization tools in a user friendly manner. NEOS relieves the end users from "software version updating" headaches and provides very friendly and expert guidance on their problems. I thank you for an excellent service.

I require students in my senior project course to study OTC thoroughly and solve some optimization problems making use of the NEOS services. These problems can be solved with the software available in our university, but I do this in order to provide students with some experience on use of Internet for long distance technical collaboration. I also believe resources like OTC are excellent means for continuing professional development for my students.

I am using NEOS for solving problems and examples for the classes I teach.

Faculdade de Ciencias da Universidade de Lisboa, DEIO

Thank you very much for BonsaiG. I used it for my master's thesis at the University of Twente, where I scheduled iterative algorithms on multiprocessor systems.

We are using the NEOS Server to design communication networks and system state perception of control agents for large, distributed networks, such as the power grid, at Carnegie Mellon University.

We are using the NEOS Server for teaching AMPL and SIF and solving tutorial examples of nonlinear optimization. Some students are using the NEOS Server for solving optimization problems when writing their Diplom-thesis.

Thank you very much for managing the NEOS-Server.

Thank you so much for your works at the NEOS Server. We are using the NEOS Server for the research of Computer Network and solving the Network Design Problem at Osaka University in Japan.

I used NEOS for a project in a grad-level CS class at Purdue Univ this past fall, on a network flow problem.

I had students run some examples using NEOS in three classes since you first announced it. No one complained :-) The only difficulties I heard about we of the i've-never-programmed-in-Fortran variety, and those are not your issue.

Valuable service.

We are planning to use NEOS with the (still experimental) callback features in inverse problems arising in multiphase reactionary fluid systems in an interdisciplinary research project at the University of Technology Aachen.

The only computation I have done using NEOS is for unconstrained NLP (NMTR), application to train an artificial neural network (in a project with an engineering student). I have also presented NEOS to third year undergraduates, to introduce them to optimization software for NLP especially, and referred 3rd parties to the NEOS web site. One of the main benefits of NEOS from my point of view is that it provides a kind of one-stop-shop for people tinkering with optimization models/methods, and is friendly enough for non-optimizers to get along with.

Education: Mathematics

I am a graduate student at the University of Pittsburgh. I am using the Filter solver (Ampl input) to find nonlinear models for the boiling points of alkanes. This is part of my Ph.D. thesis and a paper we are writing that has been accepted for publication in the journal Match. We are planning to cite the website in our paper. Dr. Mihai Anitescu is serving on my thesis committee and directed us to your site.

I am a graduate student at the UC Davis applied math program, working under the supervision of Dr. Naoki Saito. It has been our pleasure to discover the power of NEOS. Currently we are using it to solve the sparse representation problem that comes up in neuroscience (study of the brain's representation system):

I'm a PhD Student here in Spain. I'm frequently using the NEOS solvers since the beginning of this year, when I was told that they could be useful in maximizing a likelihood function which has not been implemented in any statistical package, so far. Indeed, they are very USEFUL and help me a lot. Colleagues already consult me in order to get to know the possibilities you offer via the internet. For the moment, I normally use the solver SNOPT for my programs written in AMPL, but I have to admit that, since [I am] a statistician, I need to reinforce my knowledge about the other solvers you offer.

Summa summarum: THANK you very much!!!

�� I am a part-time graduate student at the University of Southern Mississippi.� My research area is in solvers for systems of algebraic nonlinear equations.� I first used MINPACK� and have added LANCELOT (SIF), LANCELOT(AMPL) and MINOS(AMPL) to the solvers included in my studies.� I have had success with LANCELOT in solving a problem that I could not with other solvers and� I also use your solvers to verify the correctness of algorithms and other solvers that I am testing.� I constantly use your NEOS Guide and the links that your sites provide have been very useful in all of my studies.� I really appreciate the work that you have put into making your site extremely user friendly and you sharing all of your technical expertise.

Using the NEOS server in a class on Methods of Applied Math (Math511) at the Math.Dept., Univ. of Tennessee. It has been very useful. Keep up the good work!

We are using MINLP on the NEOS Server in a class on discrete design and optimization at University of Michigan. I cannot run the class without it!

We used the NEOS server for linear programming class at the University of Wisconsin-Milwaukee.

I used the NEOS server as part of my course "Computational Methods in Analysis" in both Fall 1998 and Fall 1999. There were 78 and 43 students enrolled, respectively. NEOS was used in a project to explore and evaluate methodologies for "delivering" math software to users. The students found that using the NEOS server was quite easy (although there are always a few people who seem never to be able to follow instructions) and they rated it as one of the more efficient ways to access existing software for the optimization problem they were assigned.

We are using the NEOS Server in a class on linear and nonlinear programming at Ohio University.

We have used the NEOS Server in a class for deterministic optimization in mathematical modeling at Saint Peter's College, NJ. I also use it in an elementary finite math class as a lecture demo.

I am teaching Optimization techniques for students who want become Diplom-Mathematiker. Our University is a University of Applied Sciences and for our students it is important to formulate and to solve optimization problems and it is not important to create new optimization techniques. Therefore I give only an overview about the different optimization techniques and then my students have to solve a wide variety of different optimization problems; linear and nonlinear.

First the students are learning the MPS description of linear problems and then the are learning SIF. On our computers we have an old release of LANCELOT but I ask the students to solve the problems by using the NEOS server. Sending a problem to the NEOS Server, we usually received an answer in only few minutes. This is very good! I was also interested that my students use the helpful comments and suggestions on the various NEOS Web pages. The material is very good. My work is to be a moderator between the NEOS Server and my students - I think, the NEOS Server is a good [foundation] for "Computer Based Training".

Education: O.R. & Econ.

We are using the server to give our MBA-students a problem oriented introduction to optimization. The students build small AMPL models and let NEOS do the computation. It is an extremely useful service for us, because we do not have to install any software in our computer labs.

Thanks for the prompt reaction! I am currently developing a 100% Java application to aid students to learn operations research. Since the NEOS server is by far the best place on the web where my students can submit their applications with the hope of getting back trustable results, I am planning to integrate the communication module.

Lately I've been using the Neos server to solve large optimization problems formulated as semidefinite programs. Essentially, is a combinatorial problem that models the scheduling of thermal and hydro energy resources. This is part of my research work at the University of Waterloo in Canada. Because of the size of the test sytems I've used, the neos server has been quite useful.

In my microeconomics principles class, I spend some time on cost minimization problems.� Solutions to these problems are nicely illustrated by Neos.� My students and I all appreciate having Neos available to us.

Just a comment: you are doing an excellent job and an extraordinary service to the academic community. I am regularly suggesting my students to use NEOS as soon as their projects in AMPL cannot be solved with the student edition. So they debug their AMPL models locally with AMPLPlus (which unfortunately is no longer supported from ILOG - this is a great damage to the operations research community) and then they run their real-life projects thanks to NEOS. In my laboratory we have full versions of AMPL, Cplex and MINOS, both on a PC and on a SUN server, but the students decidedly prefer to use NEOS from their lab or home PC.

So: many many THANKS!

I am a great fan of NEOS, and try to spread the good news whenever I can. My most recent success was for an economics grad student from Oxford, whose thesis was to do with multi-person co-operation. He had a conjecture that something (I can't claim to understand the details!) that was true for two-person co-operation, might not be true for three or more. He managed to formulate this as a nonlinear program, that is if the NLP had a certain solution, his conjecture was true (the NLP was actually the discretized version of a continuous problem, and he wanted a fine discretization => a large NLP). Anyway, he struggled on his own ... until I pointed him to NEOS. Within a few hours, he had formulated his problem in AMPL, sent it to a variety of solvers (LANCELOT, MINOS, KNITRO, LOQO and SNOPT), interpreted the results, et voila, his conjecture was true. I believe he has just successfully defended his thesis!

Quite apart from the fact that NEOS is extraordinarily useful for people wanting to solve real problems, it is equally wonderful for those of us who try to design new algorithms. I adore the gladiatorial aspect, put your solver up on NEOS, and let it fight, naked, against the world's finest (of course, I can't pretend that the humiliating defeats for LANCELOT don't hurt, but the experience is invaluable for our next-generation solvers).

I would like to thank the organizers of Neos for the wonderful construction which they have undertaken and which I find extremely efficient.

My students, all fourth year majors in OR take my course in Mathematical Programming and are using Neos to familiarize themselves with the best software available, which is invariably on your site. Thus, I would judge, but you should have machine data on this, that the favored software is Lancelot for constrained problems and CGplus, NMTR for the unconstrained case, the latter being a lot faster. The input into Lancelot which requires elemental and group function decomposition is a bit troublesome and it would be preferred to submit the functions in natural form. Probably this can be done, but we may have overlooked something.

We are also using Neos for research and for thesis development. This is very important to us, as we can use the abundant routines given to try and document how the other routines fare, as you will notice, from our user statistics. In this we have a bit of a problem in writing one input, but we expect that the only common input to most routines is AMPL so we shall have to learn it.

We will of course give due credit to Neos as the papers that we develop appear in print, since we are extremely grateful to your institutions for such a great service to science and to research. I have neither kept a research diary regarding out use of Neos, nor a didactic diary for my course, but as I find the Neos is a great idea and extremely well implemented, I would be glad on my side to do anything that might be useful to you to help the project. Thus I would be glad to keep such diaries, if you expect ask for user accounting from time to time. I have not done it, since I was not expecting such a request, but I would be glad to do it. Otherwise I feel that my information will remain rather generic, although highly appreciative of the project.

Neos is an extremely efficient computational tool and my classes of 14 students 4th year majors on OR following the course Mathematical Programming and a graduate class use it extensively to solve problems and thus acquire solution skills. More importantly to obtain, through solving actual problems, the feel for the theoretical underpinnings.

I develop a rather formal course in Mathematical programming, in which the basic philosophy is that you need method to be sure that you are correct. Thus given a problem they must analyze it form a theoretical viewpoint and discover what properties the solution or solutions is likely to have. Then it is a blessing to have Neos to check that their theoretical knowledge was used to good purpose.

Thus I find the Neos project a very suitable and extremely efficient laboratory for practical optimization from a viewpoint of confirming and analyzing the robustness of the theoretical derivation by good lab work. The idea of laboratory diaries, which was suggested by your request, is I think a worthwhile idea and I will pursue in the future.

I am putting up the course as an interactive distance course on the web and the Neos server is ideal for practicing. The web course which I am developing is initially in Italian and relies on this type of laboratory work, which I can now execute through the neos email server.

I suppose that I can also allow the students to connect to my machine and use the telnet facilities to submit problems on neos. At the moment, if you notice the requests spring from my machine, because they use the Linux systems in the computing lab to telnet and exploit my machine. By the way, this also works perfectly. I will keep you posted on these developments as it is part of my research.

I have, as I said, a number of prototype software which I use for optimization. The programs allow me to estimate and optimise simultaneously and I solve the problems through a recursive programming approach which solves a quadratic subproblem at each iteration, by solving a linear complementarity subproblem, under trust region constraints. I have been developing it for a number of years and, if it is ever ready and sufficiently robust, I will present it to the Neos server so that it can be used by the MP community if they want to. With this routine I have solved innumerable problems, some extremely large, some combinatorial and so on. Again if anyone is interested I will be happy to send details of the publication of the papers.

The Neos server is very useful to gauge how traditional implementations fare against my developments. This is now what I propose to do as research in a big way on many practical problems and to launch some theses, so I expect to be using Neos very much.

Of course I am very interested in all applications and do some Supply Chain optimization, solve combinatorial problems by embedding them in continuous problems, but I think at the moment, the most fun work that I am doing is to use my simultaneous estimation and optimization routines to solve diagnostic and treatment problems in Medicine. This includes Histopathological diagnosis of colon cancer, osteoporosis detection and treatment, diabetes diagnosis and treatment and so on. These are always simultaneous estimation and optimization problems and thus require comparison with other routines, so that we can develop good theory and computational expertise.

I have used NEOS Server in various courses at University of Missouri --Rolla and Bilkent University and encouraged graduate students to supplement their computational work in their thesis research with various resources provided by NEOS.

During the coming Spring semester, I will be asking students to use NEOS Server in the following two courses:

• Engineering Economic Decisions
• Design of Production Planning Systems

In addition, I find NEOS Server an excellent example of an Internet tool in decision sciences.

I teach "Modeling and Optimization" in University of Valencia (SPAIN). During the academic course, we use a GAMS software to model and solve business problems. When the number of variables exceeds the student version (50 binary or integer variables), we use the NEOS Server to solve and analyze the result of our models.

I think is very interesting the possibility to use this Internet service. In my opinion this service must be continued.

I am using neos as a tool for my OR course. The homework is to use MINLP for mixed integer programming problems. All submissions last week are from my students! NEOS is a very good tool for my teaching.

Education: Engineering

Thank you very much for being allowed to use your server. I am working on my PhD in industrial ecology. Much of my work is interdisciplinary, as a result of that I work with many different people who uses various types of tools. Some of my work has been on linking LP models and Environmental input output analysis with LCA. Just recently I've been running some MIP energy technology scenarios with endogenous learning. Working with different people requires often access to various tools. Through you NEOS Server I can both run my MP-models and my GAMS models. �Last week my solver would not handle the size of the problems I had modeled, and thanks to your server I have been able to complete my work. �

I use the NEOS server for academic purposes, as I work in the Department of Structural Engineering, Technical University of Milan, Italy. In my research activities I have encountered some challenging problems of Mathematical Programming. In detail, I am dealing with ultimate limit-state analyses of large dams by means of a non-standard approach ("direct method"); this requires solving problems of linear and non-linear programming. The NEOS server is an extraordinary tool to perform parametric tests on small models, in order to choose (also with the kind help of NEOS operators) the best suited solver. So far I have used CONOPT, MINOS and SNOPT, by way of the GAMS interface, for my non-linear, non-smooth, non-convex (very cumbersome!) problems, achieving very good results.

It was a very good experience working with the NEOS Server. I would highly compliment the stability of the server. In the one year I used the WWW version of the server, it probably crashed twice. I am indeed impressed.

I had been using XPRESS-MP MILP solver to obtain power optimal equivalent inputs to logic circuits. I work in the field of Low Power VLSI Design. I formulated large circuits as ILP problems, submitted them through the WWW interface and obtained good results.

In the later part of my research, I had to solve a large number of similar problems. So I used the mailing-in option of the server. I had written a shell script to mail my problem formulation as .mod files to the server and waited for the results to arrive back, which I parsed from the inbox input and processed.

On the whole, the XPRESS-MP solver as part of the NEOS Server was indeed a pleasure to work with.

A special mention to the tech support behind the NEOS server. I received outstanding support from them.

I have been using MINLP (AMPL version) at NEOS for my graduate class "Discrete Design Optimization." at ME Department in the University of Michigan. The class is offered every Fall semester (Sept -- Dec) and enrolled by 20-30 students. Without NEOS, I would be in big trouble...

I teach a sophomore-junior class required of all civil engineers. The title is Design and Planning of Civil Engineering Systems, and the topic is optimization as an approach to design. The course is heavy on linear programming, with some non-linear optimization as well. We use two optimization engines: that contained in Excel, and Neos. Our students have found Neos to be quite useful.

We are planning to test a number of problems related to Chemical Engineering Process design and synthesis (which are optimisation problems) for a postgraduate level course. The testing is with a view to incorporate the facilities of the server for the students.

Business & Finance

I used the MOSEK LP solver on NEOS (and lpl on my pc) to solve personal problems re asset allocation. I LOVE THIS!

I am using NEOS to compare the efficiency of my code with some codes, such as LOQO, LANCELOT, SNOPT, MINOS, .., when they solve nonlinearly constrained nonlinear network problems. It is a researching work for me.

NEOS provides an interface to a variety of solvers. We use it to evaluate various solvers/algorithms on test problems of interest. The use is both educational and commercial. Solvers used are MOSEK, XPRESS-MP for LP/MILP.

I am using NEOS servers for benchmark testing on one of the nonlinear optimization solvers. This is an important part of the decision process if we want to adopt the new optimizer. NEOS is very user-friendly, dependable and saves a lot of research time. I would continue to use NEOS server for this purpose.

I appreciate all the work you have done to make this project available for use to the public.

Insightful Corporation

We are using NEOS services for duty-scheduling for ground handling activities in an regional airport environment.

At this moment, duty-scheduling is a manual process. We are investigating how much can be gained in terms of "doing the same work with less employees" by using a more advanced scheduling mechanism. In the investigation stage "LP on demand" offers a very good solution to see how much can be gained. If the results are positive, we will consider buying software tools.

We are using the XPRESS-solver.

I work at the Swiss Federal Institute of Technology in Zurich, at the Department of Agricultural Economics. The model I have been developing includes external costs of soil erosion and water pollution in farmers' optimisation problem. The model is dynamic, spatially differentiated and includes some non-linear restrictions and binary variables. I have been using the MINLP (AMPL input) solver. It is the best non-linear solver that provides integer results I have found so far.

I'm writing a business app with an optimization sub-problem. I've formulated the optimization sub-problem as a binary LP. On my 192MB ram PC, lp_solve handles up to 10,000 variables. I'm using NEOS to try out other solvers.

I am trying to fit a nonlinear model that estimates "Value at Risk" of mexican financial institutions based on their assets, liabilities and mean duration of both.

We are using the NEOS server for making an optimal duty-schedule for employees involved in ground-handling activities in a Airport environment. For investigating the feasibility of optimization techniques we find the NEOS services very useful.

We are using the NEOS Server for verifying an algorithm designed for solving a special 0-1 ILP problem. You can find our 0-1 ILP Model in the following reference:

F Ghasemzadeh, N Archer and P Iyogun; A zero-one model for project portfolio selection and scheduling, Journal of Operations Research Society(1999) 50, 745-755

The application is a DSS based on this model.You can find more info in two last sections of the following paper and reference 3 of it.

I have modeled a seat assignment problem into the min-cost flow problem and submitted data to the NEOS solver.

Cisco Systems

We have been doing research on System Identification techniques via an NLP based maximum entropy formulation. We use system i/o data to determine the underlying parameters for the following LTI system:
x(t+1) = Ax(t) + Bu(t)
y(t) = Cx(t)
In our case, the u's represent new customers for telecommunications services. The y's represent their observed state. We seek to determine A,B,C, and x to forecast future purchasing behavior from knowledge of u and y.

The NLP formulation was one of 4 techniques we examined. We found that the formulation did not work too well for our problem. The main reason was that some of the matrix parameters must be quite small, making the problem difficult for most solvers. The equality constraints also caused problems in finding a feasible solution. We found that only LOQO would provide successful runs. Also, the predictive capability of the solution turned out to be not as good as some other techniques we examined. This result is probably application specific. I suspect applications with less extreme requirements for parameter values might work quite well.

We are using the NEOS server for testing our NLP formulations for generating financial scenarios and for testing the performances of different NLP solvers.

I use this site to run optimizations of selections of prospects to solicit for refinancing mortgage loans.

I typically mail 770,000 - 800,000 offers per week from a weekly available to mail universe of 1.7MM-2.0MM consumers. I seek to maximize the expected revenue from the mail stream by combining response rate and revenue generated.

I am self taught in AMPL and quickly outgrew the student's version which came with Fourer's terrific book. I really appreciate having the horsepower available to run my optimizations. I now complete in 5-10 seconds what took one of my subordinates 3-4 days each week to do manually.

Keep up the good work and PLEASE PLEASE keep this site open to the public.

Science & Industry

Your NEOS server has been of great help for us, We are working on bi-dimensional modeling of earth's conductivity distribution. We are testing new modeling techniques based on quadratic and also linear programming. We are a research center at Ensenada, Mexico, and our work is on applied research and education. Currently we use Knitro with the AMPL language.

I have extensively used your solver PCx to answer the question: can a linear combination of knowledge-based potentials be used as a fitness function in a genetic algorithm to predict the native folding of a globular protein? The NEOS server is very useful because it allows to save the time necessary in programming optimisation methods.

I have used NEOS with LOQO solver to optimize an interpolator. An interpolator is a device which is used to estimate the values between two adjacent samples. My domain is digital receivers where the receiver clock is not changed to match the transmitter clock so one needs an interpolator to find the necessary in-between samples. The interpolator will be used in a fractional delay filter. Normally delay filters work with integer delays; for my purposes I need to be able to delay by 0 < x < 1. This is for R&D purposes [relating to LAN/WAN systems].

I've been using NEOS lately to compare performance of various NLP algorithms that utilize second order information. We're a business of course, but this usage is a mixture of commercial and research work.

I've been testing Knitro and filter and will use other NLP codes as time and bandwidth permit.

Thanks for providing this useful service.

I have been using NEOS for circuit optimization problems involving large number of complex constraints, and my experience with NEOS is a very pleasant one. In particular I have used MINLP and SNOPT. The constant monitoring of the optimization problems being carried out is noteworthy and I especially thank the staff for killing my jobs from time to time and for making Kill Job facility available to users.

Our company is working with various projects concerning R&D of internal combustion engines for cars and brakes for heavy vehicles.

For the past several years I have been working on a system for protein structure prediction. As part of my system, I had need to incorporate a nonlinear programming solver to handle packing of sidechain atoms in the protein.

I modified my code to output packing subproblems directly in AMPL, which I could then submit to NEOS. In this fashion, I was able to evaluate solutions provided by different nonlinear solvers (I tried DONLP2, LOQO, FILTER, KNITRO, and LANCELOT), in effect testing the code before deciding which to acquire and attempt to link into my system.

NEOS is a great resource for people like me. Keep up the good work!

I have been using NEOS for circuit optimization problems involving large number of complex constraints, and my experience with NEOS is a very pleasant one. In particular I have used MINLP and SNOPT. The constant monitoring of the optimization problems being carried out is noteworthy and I especially thank the staff for killing my jobs from time to time and for making Kill Job facility available to users.

I wish to extend my regards to your extraordinary work. I have just gone through the text associated with the site and I found it very interesting and so much related to my work. I should have told you that I'm an electrical power engineer involved in the optimization of medium and high voltage power systems. I build mathematical models with either technical, economical, or hybrid�objectives and use numerical optimization algorithms to find the optimal design parameters of power systems.

NEOS is really cool! I am a die-hard GAMS user and would like to see GAMS interfaces to NLP and MINLP solvers, as my interest is primarily in solving nonlinear chemical engineering problems (production planning, scheduling, etc.) Keep up the good work!

Just discovered your great site. I have a problem arising in biophysics in which I need to find the center of a polytope generated by an LP problem (or equivalently all the vertices).

I am using the LOQO solver and code written in AMPL to perform numerical optimization of a spinor Bose-Einstein condensate. This yields the ground state energy of the atomic system and the corresponding wave functions as well as a variety of other properties of superfluids. The aim is to develop a fundamental theory which can explain the recent observations of Bose-Einstein condensates.

I have used NEOS as an excellent guide to optimisation methods and locating associated software. Its the best resource I know of on the web and has been very useful in the development of optimisation methods in our process modeling software. Therefore, I welcome any developments to improve the site.

AspenTech Ltd

We believe the NEOS optimization server provides a very valuable service.

ABAQUS/Design Development Manager
Hibbitt, Karlsson & Sorensen, Inc.

We are using the NEOS Server for solving linear and nonlinear complementarity problems in engineering mechanics and in robotics (till now, mainly unilateral contact problems and friction).

The problem we are considering is the design of a yagi antenna. These antennas consist of an array of wires and are frequently used as TV antennas. In the 1960s a lot of research has been done on the best design of this type of antenna. However, all research was experimental and therefore very costly and time consuming. Our idea is trying to design antennas by using the computer as a tuning device, i.e. starting from an initial design to have the computer finding the optimal solution by using constrained optimization methods. If this approach works, it could be used for the design of the future Square Kilometer Phased Array (SKA) for radio astronomy to look back in time to the first moments of the universe, which is to be build in about 10 years from now.

We have tried various solvers on NEOS to see if this is possible at all and the yagi antenna has been used as a test case for which good validation is possible. The NEOS server offers a lot flexibility and various (well-known) solvers. It gives us the opportunity to find out which approach works best. Without this server it would probably be impossible to find out what strategy is most likely to succeed. Preliminary results are very encouraging for this approach.

I would like to express my support for this service. I am using this server to solve a non-linear optimization problem to determine the resource requirements for broadband networks using effective and decoupling bandwidths. The objective function in my optimization is assumed to be linear. The constraints are non-linear, exponential and sometimes conditional.

The NEOS server helped me tremendously, since my background in optimization is weak. I was nevertheless able to further my research using this service

I just used AMPL-MINOS to solve an LP arising from a stochastic programming application that Jeff and I are looking at. The debug-resubmit process could not have been simpler nor more convenient. Fine stuff.

Here at our group at the Philips Research Laboratories we have used the NEOS Server in our search for nonlinear optimization methods for circuit simulation purposes.

I just found NEOS while looking around for a solution to my problem (forgive the pun) attempting to optimize a couple of functions for a computer game I am designing. It's very helpful, thank you

I am using the NEOS Server to solve optimization problems that arise in (1) distance geometry and multidimensional scaling; and (2) design of experiments.

We have used the NEOS server to solve NLP problems we formulated from Integrated Circuits Computer Aided Design. And we plan to use it again in the near future.

I have been using your AMPL front end and solvers to develop an assignment problem. BTW, this is a great service; allows me to develop without incurring client costs or requesting software my university cannot afford.

I used your codes (Minpack) in the past. Any deployment of nonlinear methods on a server are beneficial and their importance will increase.

Thank you for your help. NEOS is very useful to me : I am solving huge multi-commodity flow problems written in AMPL, and we do not have AMPL available here. I have been using NEOS for solving multi-commodity flow problems arising in the optimization of weights for routing in the Internet.

Congratulations for this nice tool and your continuous efforts to improve it.

I did send problems to the PCx and the BPMPD solver and was very pleased with the outcome. These were rather small problems in the field of chip design. Larger problems consumed too much computer memory, let alone the time a solution might have taken.

Nevertheless, information is well categorized, easily available and very supportive, as well as the people involved are. I had misunderstood the right-hand-side labels of the MPS-format. Not realizing that they supported different versions, I thought each RHS had a different label. I reported the resulting output as not good and by the time I noticed my mistake, one of you pointed it out to me as well. Good luck with your efforts in making the NEOS server even better.

I have used some NLP solver from NEOS to solve a circuit optimization problem. It's very helpful to use it this way, other than installing those packages by myself.

By way of introduction let me mention that as a sideline to my professorial duties, I maintain a web site which seeks to direct people to appropriate parts of mathematics; my intended audience is students and professionals who have at least a couple of years exposure to college level mathematics.

Turn to http://www.math-atlas.org/welcome.html and you'll see a set of blobs on the screen; there's also one on each index page e.g. http://www.math-atlas.org/index/15-XX.html . The positions of these blobs were determined by the NEOS server. I was happy to have access to the server because I think having the blobs positioned appropriately helps understand the topic, and therefore adds to the quality of the web pages.

What these blobs are is sets of papers published in mathematics (specifically, papers reviewed in Mathematical Reviews 1980-1999). Each paper in MR is assigned one or more classification codes from a subject classification system of mathematics. I used the incidence data showing how many papers published in various sub-disciplines were given secondary classifications in other sub-disciplines; these data gave a sort of metric on the subject areas, which indicates that say Group Theory is close to Ring Theory but far from Differential Equations. So conceptually I then have the various disciplines arrayed in some high-dimensional space, some pairs of them closer to each other than other pairs. In my mind I visualize the subjects hovering around like stars in the Milky Way galaxy.

Well, there are standard tools for projecting such a picture to a plane so as to show the distance relationships optimally -- that is, you can sort of figure out the "main plane" of the galaxy. The problem is that when you project the various blobs to the plane of the galaxy, some of them will obscure others which are directly above or below them.

So here's where I used NEOS: I took the nominal positions of the different blobs and asked for the minimal amount of jiggle I would need to separate them on the plane. It's not a hard project, conceptually, and I imagine cartographers have packaged routines to deal with this. But certainly there were too many variables and constraints to allow an analytic solution (and way too much computation to appeal to a theoretician like me). It took me a little while to master the NEOS syntax (e.g. I had to remember to tell it to display the answer...) but once I got it down to a science I was able to have the server supply the results I wanted in a painless way.

There's quite a bit of arbitrariness in the way the data are massaged to produce a picture: one shouldn't try to read too much into these images. On the other hand, when left to their own devices, the programs tended to draw pictures which corroborated what I would have drawn on my own (when I had some prior knowledge of a sub-discipline) and gave hints about what real connections existed in sub-disciplines with which I was not familiar. So I'm quite happy with the results so far.