About the Journal

The Journal of Computational Optimization and Scientific Machine Learning (JCoSML) is an international, peer-reviewed, diamond open-access journal publishing original research at the interface of mathematical optimization, numerical analysis, and machine learning.

Focus and Scope

The Journal of Computational Optimization and Scientific Machine Learning (JCoSML) publishes original research at the interface of mathematical optimization, numerical analysis, and machine learning. The journal is concerned with methods that are both computationally practical and analytically understood, and it welcomes work that strengthens either side of that pairing.

Topics within scope include, but are not limited to:

continuous, discrete, and stochastic optimization: first- and second-order methods, variance reduction, distributed and decentralised algorithms, derivative-free and Bayesian optimization, metaheuristics and evolutionary computation, convex and non-convex programming, multi-objective optimization;
scientific machine learning: physics-informed neural networks, neural operators, surrogate and reduced-order modelling, differentiable simulation, hybrid numerical–learning solvers;
optimization and computational methods for large language models and foundation models: training and fine-tuning algorithms, parameter-efficient adaptation, quantisation and model compression, inference-time scheduling and serving efficiency, distributed and memory-constrained training, scaling behaviour, and the theoretical analysis of these methods;
numerical methods for differential equations where a learning component plays a substantive role;
inverse problems, uncertainty quantification, and data assimilation;
theoretical analysis of learning algorithms: convergence, generalisation, complexity, and stability;
optimization in machine learning practice: training dynamics, hyperparameter optimization, neural architecture search, model compression, and federated or privacy-preserving formulations;
data-driven decision making and operations research: scheduling, routing, resource allocation, and combinatorial problems addressed by learning-augmented methods;
applications in engineering, physics, energy systems, computational biology, and the geosciences, where the contribution lies in the computational or mathematical method rather than in the application alone;
high-performance and hardware-aware implementations of the above, including reproducible software contributions and benchmark infrastructure.

Purely empirical benchmarking studies without methodological or analytical contribution, and applications of standard machine-learning pipelines that raise no computational or mathematical question, fall outside the scope.

Open Access Policy

JCoSML is an open-access journal: all content is freely available without charge to the user or the user's institution. Readers may read, download, copy, distribute, print, search, link to the full texts, and pass them to software as data, without prior permission from the publisher or the author. This corresponds to the definition of open access set out in the Budapest Open Access Initiative.

The journal operates a diamond model: there are no submission fees, no article processing charges, no page or colour charges, and no fee of any kind is requested at any stage. Publication costs are borne by the publisher, Salih Berkan Aydemir . Any message purporting to come from JCoSML and requesting payment should be reported to the editorial office.

Plagiarism Policy

All manuscripts submitted to the Journal of Computational Optimization & Scientific Machine Learning are screened with iThenticate before peer review. Submissions with an overall similarity index above 25%, or with more than 5% overlap from any single source, are returned to the authors without review. The reference list, standard methodological phrasing, and correctly quoted and attributed material are excluded from this calculation.

A similarity index below the threshold does not by itself establish originality: the editor examines every flagged passage individually, and text that reproduces another work's substance without attribution is treated as plagiarism regardless of the reported percentage. Self-plagiarism, including substantial reuse of the authors' own previously published text without citation, is held to the same standard.

Manuscripts found to contain plagiarism after acceptance are retracted, and the authors' institutions are notified. The journal follows COPE guidelines in handling all such cases.

Copyright Notice

Authors retain copyright in their work without restriction. Publication in JCoSML involves no transfer of ownership. On acceptance, authors grant the journal a non-exclusive licence to publish, archive, and disseminate the article, and agree to its release under the Creative Commons Attribution-NonCommercial 4.0 International licence (CC BY-NC 4.0).

Under CC BY-NC 4.0, anyone may share and adapt the article for non-commercial purposes, provided that credit is given to the author(s), the article title, the full journal citation, and the DOI, that a link to the licence is provided, and that any changes are indicated. Commercial use requires prior permission from the copyright holder. Articles may be used for text and data mining for non-commercial purposes subject to these conditions. Third-party material reproduced in an article remains subject to the terms of its own licence and is excluded from the article licence.

Authors may deposit the submitted, accepted, or published version in any repository or on any website at any time, without embargo

Peer Review

Submissions are first checked by the editorial office for scope, completeness, formatting, and textual overlap. Those that pass are assigned to a handling editor and sent to at least two independent reviewers holding a doctorate in the subject area, under single-anonymised review: reviewers know the identity of the authors, authors do not know the identity of the reviewers. Reviewers are given fifteen days to report, with one extension available. The handling editor recommends a decision, which is confirmed by an editor-in-chief. Typical intervals are two weeks to a first editorial decision and six to eight weeks to the first reviewers' reports; the complete process ranges from one to eight months depending on reviewer availability. The full workflow, including the decision rules, is published separately.

Publication Ethics

JCoSML follows the Core Practices of the Committee on Publication Ethics (COPE). Authorship follows the CRediT taxonomy and requires a substantial contribution together with accountability for the work as a whole. All submissions are screened for textual overlap. Competing interests and funding must be declared, and a data availability statement is required.

Generative artificial intelligence cannot be listed as an author. Its use for language editing, code assistance, or literature search must be disclosed; authors remain fully responsible for every claim, proof, and citation in the article. Reviewers and editors must not upload manuscripts to any AI service, as this breaches confidentiality.

Where a member of the editorial board submits to the journal, the manuscript is handled by an independent editor with no access to the identity of the reviewers, and that arrangement is stated in the published article.

Archiving and Preservation

This journal utilizes the PKP Preservation Network (PKP PN) to create a permanent, distributed archive of its published content for the purposes of preservation and restoration. The PKP PN uses LOCKSS (Lots of Copies Keep Stuff Safe) technology to ensure that all published articles remain accessible in the long term, even if the journal ceases publication.

In addition, all articles are assigned DOIs registered with Crossref and are self-archived on the journal website.

Publication Frequency

JCoSML publishes two issues per year, in June and December. Accepted articles appear in early-view form with a DOI as soon as they are approved by the editor-in-chief, and are assigned to the first available issue. Early-view articles are citable from the moment of publication.

Sponsorship and Support

JCoSML is published by Salih Berkan Aydemir. The journal carries no advertising, accepts no sponsored content, and receives no commercial funding. Its costs are met by the publisher, and no charge is made to authors or readers at any stage. Editorial decisions are made solely on scholarly grounds and are independent of any financial consideration.

Privacy Statement

The names and email addresses entered in this journal site will be used exclusively for the stated purposes of this journal and will not be made available for any other purpose or to any other party. Reviewer identities are not disclosed to authors, and reviewer reports are not published.

Contact

Editorial Office — Journal of Computational Optimization and Scientific Machine Learning
Publisher: Salih Berkan Aydemir, Amasya, Türkiye· salihberkanaydemir@gmail.com · admin.jcosml@gmail.com

About the Journal