The theory and practice of computing and operations research are necessarily intertwined. The INFORMS Journal on Computing publishes high quality papers that expand the envelope of operations research and computing. We seek original research papers on relevant theories, methods, experiments, systems, and applications. We also welcome novel survey and tutorial papers, and papers describing new and useful software tools. We expect contributions that can be built upon by subsequent researchers or used by practitioners.

Areas and Area Editors

Editorial statements for each Area are listed to guide authors in the selection of an appropriate Area for submission.

Applications in Biology, Medicine, & Healthcare
Computational Modeling: Methods & Analysis
Data Science & Machine Learning
Design & Analysis of Algorithms - Continuous
Design & Analysis of Algorithms - Discrete
Heuristic Search & Approximation Algorithms
Network Optimization:  Algorithms & Applications
Simulation
Software Tools
Stochastic Models & Reinforcement Learning
Applications in Biology, Medicine, & Healthcare

J. Paul Brooks
Supply Chain Management and Analytics
Virginia Commonwealth University
Richmond, Virginia, USA
jpbrooks@vcu.edu

The frontiers of biological, medical, and healthcare research depend on sophisticated modeling, analysis, and computational techniques. Operations research, including all aspects of optimization, stochastic processes, and simulation, are vital tools for investigating complex biological systems, advanced medical procedures, and healthcare delivery processes. In turn, the computational challenges associated with these applications have spawned new developments in operations research, creating a synergy of application, theory, and implementation.

Submissions that include advancements in OR methodology for addressing a problem in biology, medicine, and/or healthcare are welcome. In addition, we welcome submissions presenting innovative applications of existing OR methodologies to address new problems in their respective domains. Submissions should include aspects of biology/medicine/healthcare, operations research/management science, and computer science/computing. Accepted papers will provide a significant contribution in one of these three aspects or possibly provide a contribution in some combination of the three. Accordingly, reviewers may be selected from multiple disciplines to provide complementary perspectives in evaluating submissions.

Examples of computational OR applied to biology/medicine/healthcare:

    computational OR methods for improving healthcare delivery
    development of OR-based machine learning algorithms for biomedical data
    scheduling of healthcare workers and patients
    advancements in the use of optimization for radiation treatment therapy
    modeling of infection control and antibiotic stewardship programs
    the use of OR methodology supporting the establishment of healthcare policy
    novel OR-based algorithms for computational biology

Computational Modeling: Methods & Analysis

Pascal Van Hentenryck
H. Milton Stewart School of Industrial and Systems Engineering
Georgia Institute of Technology
Atlanta, Georgia, USA
pascal.vanhentenryck@isye.gatech.edu

Rapid advances in computing technology have permitted the solution of increasingly large and complex operations research models. As a consequence, much greater attention must be paid to formulation, modeling, and computational issues. This includes the need to understand how to model complex problems effectively, how to compare formulations, and how to find the proper tradeoffs between model fidelity and the required computational resources to solve it. The area also covers the need for novel techniques and tools for acquiring, formulating, debugging, analyzing, and visualizing models. Modeling Methods and Model Analysis are the two main overlapping themes in this very broad area:

    Modeling Methods covers research on ways of creating and managing operations research models which feature a significant computational component. It includes computational techniques to model uncertainty, represent complex data sets and solutions, and capture complex decisions, objectives, and constraints. It also covers techniques to acquire constraints and elucidate preferences, algebraic modeling languages, and graphical representations. Additionally, this theme includes how to formulate and compare models for highly complex applications, trading off model fidelity and computational requirements. Optimization for novel computational paradigms such as quantum computing is also considered.

    Model Analysis covers research on ways to analyze models and model instances to provide useful insights. Examples include computational methods and tools for explaining, debugging, visualizing, and analyzing models. It also covers techniques for reformulating and simplifying models, and for infeasibility analysis.

Data Science & Machine Learning

Ram Ramesh
Department of Management Science and Systems
School of Management
SUNY at Buffalo
Buffalo, New York, USA
rramesh@acsu.buffalo.edu

In our increasingly information-based economy and society, data-driven decision-making and data-driven automation have become central to most human endeavors. Data Science is rooted in the domain of data-driven decision-making and Machine Learning is centered on data-driven automation. Data Science spans the four dimensions of the analytics field – diagnostic, descriptive, predictive, and prescriptive analytics. Similarly, Machine Learning is focused on modeling data, learning from data, and applying them in the automation of processes. Although their target application domains could be different, Data Science and Machine Learning employ similar models, algorithms, and systems, and both are well established in the underlying disciplines of statistics, optimization, and computer science. A common thread that unifies the two fields is their consistent thrust on extracting knowledge and insights from possibly noisy data, and applying this knowledge and data-supported insights to solve both structured and ill-structured problems in a variety of application domains. The explosive growth in the fields of statistical modeling, optimization, algorithm design, and computing systems have greatly enhanced the scope and reach of Data Science and Machine Learning in contemporary economy and society.

The Data Science & Machine Learning (DSML) area provides a forum for publication of papers on the cutting edge of these rapidly evolving disciplines. The area welcomes innovative contributions with a strong emphasis on analytical modeling, algorithmic development, and computational experimentation leading to validation. Given the multi-disciplinary scope of the journal focusing on the interface of computer science, information systems, and operations research, topics that are most appropriate include:

    Data modeling and learning approaches that demonstrate significant performance or functionality improvements
    Methods for efficiently managing distributed data or knowledge
    Innovations in the use of DSML methodologies in important application domains (to name a few, consumer behavior, preference analysis, social media, financial technologies, GIS, agent-based applications)
    Innovations in DSML methodologies leading to novel and superior solutions to both structured and ill-structured problems (to name a few, data mining, link prediction, knowledge discovery, graph modeling, topic modeling, statistical modeling)
    Advances in methods for analyzing performance of DSML systems

Design & Analysis of Algorithms - Continuous

Antonio Frangioni
Department of Computer Science
Università di Pisa
Pisa, Italy
frangio@di.unipi.it

Design & Analysis of Algorithms - Discrete

Andrea Lodi
Department of Mathematical and Industrial Engineering
Ecole Polytechnique de Montréal
Montréal, Canada
andrea.lodi@polymtl.ca

The Design & Analysis of Algorithms Areas seek to publish significant contributions to the algorithmic aspects of operations research. The development of algorithms for optimization problems is a vibrant field due to the need, on one hand, to develop general-purpose solvers capable to tackle as large as possible classes of optimization problems, and, on the other hand, to exploit as much as possible the structure of the problem at hand to improve the effectiveness and efficiency of the approach. This is especially relevant when the problems are of large scale and/or tight constraints are imposed on the available computational resources, as it happens in many applications. Although a single objective function is most often considered, solutions methods for multi-objective optimization problems, equilibria/variational ones and games are also welcome.

Thus, the design of algorithms covers a broad spectrum of issues, ranging from the study of the complexity or approximability of a problem to an algorithmic engineering project involving high-performance computing platforms, advanced data structures, and real-life data. The analysis of algorithms may concern all the nontrivial aspects related to their completeness, correctness, and efficiency as expressed by the trade-off between the solution quality they deliver and the amount of computational resources they require, estimated by either theoretical (worst-case or probabilistic) or experimental means.

In principle, algorithms covering all possible applications (resource allocation, facility location, routing and scheduling, energy, transportation, finance, just to mention a few) and of all possible types (exact, heuristic, continuous, combinatorial, etc.) are welcome in these two areas, subject to the specificities of each detailed below. However, other areas of the Journal focus on either specific applications (Biology, Medicine, & Healthcare; Knowledge Management & Machine Learning; Network Optimization) or on specific solution techniques (Stochastic Models; Heuristic Search; Simulation). In general, a research paper should be submitted to the Design & Analysis of Algorithms Area if the algorithm is designed either for a quite general class of problems (although it can be motivated and applied to specific cases), or for an application not covered by another specific area; otherwise, it should go to the more specific area.

Also, the two individual areas

    Design and Analysis of Algorithms: Continuous
    Design and Analysis of Algorithms: Discrete

have a very significant intersection, in particular that of algorithms for Mixed-Integer Non Linear Problems that inherently require to tackle both their continuous and their discrete aspects. In this case, the contribution should go to the area that, to the judgement of the authors, is the most fitting one, with knowledge that the Area Editors may override this decision.

Design and Analysis of Algorithms: Continuous

This area actually covers algorithms for three possibly different kinds of approaches:

    exact algorithms for continuous problems whose global optimality can be "easily" demonstrated, typically convex ones;
    local algorithms for (nonconvex) continuous problems;
    exact algorithms for nonconvex continuous problems whose global optimality is "hard" to establish.

Thus, different aspects can be at play in each case, such as algorithms for very-large-scale convex and even linear programs (with such applications as Machine Learning, stochastic, chance-constrained and robust optimization, and others), rate of local convergence and robustness of nonlinear (nonconvex) algorithms, bounding and/or search techniques for global nonlinear nonconvex optimization. These aspects may, of course, co-exist and interact in a given application; besides, as previously noted, global optimization techniques for nonconvex problems have strong relationships with those for discrete optimization. All research work pertaining to these aspects is welcome to this area regardless of specificities of the problem at hand (class of nonlinear constraints/functions involved, availability and continuity of derivatives, ...) unless the contribution is narrowly focused on some application or methodology for which a more specific area exists.

Design and Analysis of Algorithms: Discrete

This area is the home for

    combinatorial,
    constraint-based,
    logic-based, and
    linear-programming-based

exact algorithms for difficult discrete optimization problems.

As anticipated in the part of editorial statement in common with the Continuous Area, algorithms for Mixed-Integer Nonlinear Optimization are especially welcome and the responsibility of this topic will be shared.

We are particularly interested in hybrid algorithms combining several of the above techniques to achieve computational effectiveness.

Finally, we especially value the generality of the proposed algorithm, thus the applicability of a technique within general-purpose solvers and its effective implementation. Examples are cutting plane generation, branching rules, preprocessing and filtering techniques, etc. that can be used to enhance the computational performances of general purpose methods / solvers.

Heuristic Search & Approximation Algorithms

Erwin Pesch
Economics and Business Administration
University of Siegen
Siegen, Germany
erwin.pesch@uni-siegen.de

This area focuses on the design, learning, and application of efficient and innovative methods for approximately solving relevant, difficult (combinatorial) optimization problems. In particular, it covers topics such as approximation algorithms with performance guarantee, (fully) polynomial time approximation schemes, parameterized algorithms, and heuristics. New ideas in rounding data and dynamic programming, deterministic or randomized rounding of linear programs, greedy, nature-inspired and local search algorithms, neural networks, and constrained programming-based learning, or hybrid approaches combining existing heuristic methods, alone or in conjunction with techniques from other areas of operations research or computer science, are also of particular interest. The emphasis within the area is on papers presenting methodological innovations that can be applied to a wide range of problems or situations and include a proof of performance. Papers must clearly elaborate the innovation either in modeling, or a new solution approach, justified theoretically or via computational comparison with existing methods and/or in a real application. A simple assertion that the innovations can be applied elsewhere does not meet the burden of proof required in good scientific practice. The authors must demonstrate generality in a convincing manner, either experimentally or theoretically. Rarely, it may be clear that a method is broadly important even when it is tested on only one problem, but this would be very unusual.

For many outlets in our field, a necessary and sufficient condition for publication is to show better results than a set of competitors over some test instances. While this may be a reasonable standard in some settings, it is generally insufficient for the INFORMS Journal on Computing, especially given that a significant fraction of our audience are not specialists in heuristic search and learning. Obtaining best-known results will be helpful in making the case for the paper, but it is not normally sufficient.

The problem(s) considered in the paper must be “important” in some sense, though this is difficult to define precisely and is subject to some tradeoffs. Importance may be demonstrated by application to a problem of practical significance, or demonstration that the problem has been extensively studied in the research literature, for example.

A scientifically rigorous paper presenting methodological innovations that can be applied to a wide range of important problems or situations is not easy to produce. However, it is relatively easy to recognize such a paper. Papers in this category will not generally require multiple major revisions: a paper presenting methodological innovations that can be applied to a wide range of problems or situations does not need a committee to make it perfect. Conversely, the referees generally cannot walk authors through the process of discovering and describing methodological innovations that can be applied to a wide range of problems or situation.

Survey papers covering recent advances in a given field and papers aimed at providing a conceptual integration of the area are also welcome.

Network Optimization: Algorithms & Applications

Russell W. Bent
Los Alamos National Laboratory
rbent@lanl.gov

The Network Optimization Area focuses on networks across the social, natural, physical, and engineering sciences. This focus ranges from engineered networks like electric grids, telecommunications, and transportation, to network models of social interactions, and to natural processes like biological networks. Of primary interest are those papers that combine fundamental contributions in the advancement of computational methods for network science (optimization, game theory, artificial intelligence, etc.) with impactful computational demonstrations on an application domain.

Topics of interest include, but are not limited to, the following:

    Algorithms for analysis, design, planning, and control of networks;
    Network security, survivability, robustness, risk, and resilience;
    Network applications in emerging technological areas; and
    Case studies involving real, large-scale network problems including publishable data and a computational method contribution.

Manuscripts which apply or develop graphical methods (such as neural networks) for non-network problems are outside the scope of this area.

Simulation

Bruno Tuffin
IRISA/INRIA
Campus de Beaulieu
Cedex, France
Bruno.Tuffin@inria.fr

The Simulation Area is interested in research in all computational aspects of simulation. We are looking for high-quality research on the computational aspects of:

    Simulation model building (for instance, object-oriented language development)
    Simulation data structures (e.g., specialized data structures for simulation models using historical data)
    Simulation modeling/experiment environments (such as web-based simulation)
    Simulation stochastic input modeling (like Johnson-family or Bézier distribution fitting)
    Random-variate generation (for instance, time-series random variate generation)
    Output analysis (such as optimal batch-size selection theory and algorithms)
    Optimization over small feasible sets (like ranking & selection procedures and multiple-comparison procedures)
    Optimization for large feasible sets (such as random search methods)
    Optimization over sets containing continuous variables (e.g., likelihood-ratio methods)
    Variance-reduction methods for simulation experiments over classes of problems possessing special structure (such as order(s)-of-magnitude variance reduction in Monte Carlo evaluation of financial options)
    and other aspects of simulation experiment modeling and analysis.

Tutorials and applications in the above areas that are comprehensive and integrate from multiple sources are also welcome.

Software Tools

Ted Ralphs
Industrial and Systems Engineering
Lehigh University
Bethlehem, Pennsylvania, USA
ted@lehigh.edu

Software Tools seeks papers describing software and/or data made available to the research community.

In the case of software, the contribution should be novel with respect to software, but need not necessarily be so with respect to the underlying mathematics or algorithms. The novelty could be in a sophisticated implementation that showcases state-of-the-art techniques, an ingenious abstraction that simplifies an important and previously difficult research task, or a robust implementation of an existing state-of-the-art algorithm where none previously existed. The software should be of high technical quality; follow best practices for software engineering; and be usable, maintainable, and of long-term interest to the research community. The paper should make the case for the novelty of the software, describe how it fits into the literature, and explain clearly how it can be leveraged in solving a computational problem of importance to the research community. The paper should not be a user’s manual or a technical specification.

In the case of data, the paper should address why the data is important to the research community in advancing research and what methods were used in producing it. The novelty could lie in the way that the data was collected and processed/cleaned (for example, the data may be high velocity or high volume, requiring sophisticated collection techniques), how it was generated, or how the data was integrated into a novel data analytics workflow to solve a problem of high interest. It is likely that any paper addressing data will also address associated software.

An archive of the data and/or software that is the subject of the paper is expected to be made available to the community through the IJOC Github repository (https://INFORMSJoC.github.io) following guidelines described in https://INFORMSJoC.github.io/InstructionsForAuthors.html. Such an archive should be submitted along with the paper for review. There should be some expectation that the software will remain useful in the long-term and it should be released under a license that makes it useful for research.

Both terse papers, briefly describing a software or data contribution and its importance to the research community, and full expository papers are sought. Short papers focused exclusively on the software or data will be subjected to the usual review process, but are not expected to undergo multiple rounds of revisions. Full-length papers must include additional material with enough significance to merit publication. In both cases, the review process will consist of two separate parts: a review of the software/data, to assess the quality and maintainability, and a review of the paper itself, assessing the novelty and potential importance to the research community.

Stochastic Models & Reinforcement Learning

Nicola Secomandi
Jones Graduate School of Business
Rice University
Houston, Texas, USA
Nicola.Secomandi@rice.edu

The Stochastic Models & Reinforcement Learning Area welcomes manuscripts that deal with the computational aspects of models that represent systems in which uncertainty is a central concern and methods that evaluate or optimize the performance of these systems. Computation is a critical dimension in dealing with large-scale stochastic systems that defy exact analytical solution. The challenge is to develop efficient, effective, and reliable algorithms for such systems. Papers of interest develop new such methodologies; conduct novel and insightful analysis of methods; and report on timely, important, and innovative applications based on data or realistic parameter values. The scope of this area includes a variety of topics. Examples include Markovian modeling of stochastic systems; aggregation approaches for Markov chains, such as ones that arise in data-driven applications that feature huge data sets; and reinforcement learning, approximations, and bounding techniques for intractable Markov decision processes and stochastic dynamic programs. (Papers that are primarily based on mathematical programming or focus on supervised or unsupervised learning should typically be submitted to one of the other areas of the journal.) Appealing manuscripts deal with models of both traditional and emerging contextual domains, such as inventory, supply-chain, and service management; revenue management, pricing, and market analytics; social networks; energy; and financial engineering, computational finance, and risk management. Clear and concise exposition and rigorous execution are defining elements of successful articles.