Machine Learning

《机器学习》（Machine Learning，常用缩写为 Mach. Learn.）是一份由施普林格出版的计算机科学同行评审学术期刊，月刊。该刊创刊于 1986 年，是机器学习领域历史最久、覆盖面最广的刊物之一，刊载各类学习方法应用于不同学习问题的实质性成果。Machine Learning (commonly abbreviated as Mach. Learn.) is a peer-reviewed computer science journal published by Springer, issued monthly. Founded in 1986, it is among the longest-running and broadest-scoped titles in machine learning, publishing substantive results on learning methods applied to a wide range of learning problems.

Scope

本刊的取稿立场是「以计算方法本身为研究对象」，即要求论文须在方法、理论或实验设计上有实质贡献，而非仅把现成模型用于某个新数据集。刊物同时要求结论具备可验证与可复现性，并把方法学严谨性作为独立评价维度。收文覆盖的主要方向包括：The journal's editorial stance takes computational method itself as the object of study, requiring that a paper contribute substantively to method, theory or experimental design rather than merely applying an existing model to a new dataset. The journal additionally requires that findings be verifiable and reproducible, and treats methodological rigour as an independent criterion of appraisal. Its principal subject areas include:

• 学习问题： 分类、回归、识别与预测，以及规划、推理与推断等问题求解任务。Learning problems: classification, regression, recognition and prediction, together with problem-solving tasks such as planning, reasoning and inference.
• 监督与无监督方法： 决策树与规则学习、连接主义网络、概率与统计模型、归纳逻辑编程与基于案例的方法。Supervised and unsupervised methods: decision-tree and rule learning, connectionist networks, probabilistic and statistical models, inductive logic programming and case-based methods.
• 集成与强化方法： 集成学习、聚类、强化学习、演化方法与基于解释的学习。Ensemble and reinforcement methods: ensemble learning, clustering, reinforcement learning, evolution-based methods and explanation-based learning.
• 数据挖掘与信息检索： 数据挖掘、网络挖掘、信息检索、自然语言处理与科学发现。Data mining and information retrieval: data mining, web mining, information retrieval, natural language processing and scientific discovery.
• 应用领域： 视觉与语音感知、机器人与控制、组合优化、博弈，以及工业、金融与科学应用。Application domains: vision and speech perception, robotics and control, combinatorial optimisation, game playing, and industrial, financial and scientific applications.

该刊属混合出版模式：订阅路线不向作者收费，选择开放获取需支付文章处理费。出版方官方公布投稿至首次决定中位数为 5 天，该口径包含编辑部初筛。该刊被科学引文索引扩展版收录，2025 年影响因子 4.9。The journal is a hybrid title. The subscription route charges authors nothing, while open access carries an article processing charge. The publisher states an official median of 5 days from submission to first decision, a figure that includes editorial screening. The journal is indexed in the Science Citation Index Expanded and carries a 2025 impact factor of 4.9.

History

• 创刊（1986年）： 本刊创刊于 1986 年，由施普林格出版，是机器学习领域最早的专业期刊之一，与 1980 年代中期机器学习从人工智能中独立成科的进程同步。Founding (1986): the journal was founded in 1986 and published by Springer, one of the earliest specialist journals in machine learning, and it emerged alongside the process by which the field became independent of artificial intelligence in the mid-1980s.^[1]
• 印刷与电子刊号： 本刊印刷刊号 0885-6125、电子刊号 1573-0565，两者并行使用；库内所记的创刊年与刊号体系一致。Print and electronic serial numbers: the journal carries print serial number 0885-6125 and electronic serial number 1573-0565, both in parallel use, agreeing with the library's record of the founding year and the numbering system.^[2]
• 出版方沿革： 创刊后长期由施普林格（纽约）出版，现归属施普林格自然集团，库内出版方记录为施普林格，方向正确但可补全为现属关系。Publisher history: the journal was long published by Springer of New York and now belongs to the Springer Nature group. The library's publisher record of Springer is correct in direction but can be completed with the current ownership.^[3]
• 官方审稿口径： 出版方官方公布投稿至首次决定中位数为 5 天，该口径包含编辑部初筛；库内原记「较慢，6 至 12 周」无官方依据，已删除。出版方不公布录用率。Official review timing: the publisher states an official median of 5 days from submission to first decision, a figure that includes editorial screening. The library's previous note of a slow six to twelve weeks had no official basis and has been deleted. No acceptance rate is published.^[4]
• 指标走势： 影响因子自 2025 年的 4.9 与五年影响因子 8.0 共同刻画该刊位置；中科院分区中列于计算机科学大类，小类人工智能在 2025 年升级版中为四区。Metric trend: the 2025 impact factor of 4.9 together with the five-year figure of 8.0 places the journal in the computer science class of the Chinese Academy of Sciences ranking, with the artificial intelligence sub-category at zone 4 in the 2025 upgraded edition.^[5]
• 规模与国际化： 该刊 2025 年论文下载逾二百四十万次，作者来源以美国、英国、德国、中国、日本与法国为主，国际化程度在计算机科学刊物中突出。Scale and internationalisation: the journal recorded more than 2.4 million downloads in 2025, with authors concentrated in the United States, the United Kingdom, Germany, China, Japan and France, an unusually international profile for a computer science title.^[6]

Editors

本刊实行主编负责制，主编由施普林格任命并主持编辑部，稿件经 Editorial Manager 系统投递。官方不公布除首次决定时长之外的编辑流程细节，编辑部构成以刊头署名为准。The journal operates under a single editor-in-chief appointed by Springer, with manuscripts submitted through the Editorial Manager system. The publisher publishes no editorial process detail beyond the median time to first decision, and the composition of the editorial office is taken from the masthead.

• 米开朗基罗·切奇（现任主编）： 巴里阿尔多·莫罗大学教授，数据挖掘与机器学习研究者；其主编身份在施普林格官方期刊页的编者栏中载明。Michelangelo Ceci (current editor-in-chief): professor at the University of Bari Aldo Moro, working on data mining and machine learning, whose editorship is recorded in the editors' block of Springer's official journal page.^[7]
• 学科奠基地位： 作为机器学习领域最早的专业期刊之一，该刊的创刊与该学科在 1980 年代中期的独立进程同步，其早期论文构成学科奠基文献的一部分。Foundational position: as one of the earliest specialist journals in machine learning, the journal's founding ran in step with the field's emergence as an independent discipline in the mid-1980s, and its early papers form part of the foundational literature of the field.^[8]
• 出版方关系： 刊物长期在施普林格体系内出版，创刊期由施普林格纽约分部负责，现归属施普林格自然集团，编辑部与出版方在流程上分工明确。Publisher relationship: the journal has been published within the Springer system throughout, with the New York branch responsible at the founding stage and the title now belonging to the Springer Nature group, the division between editorial and publishing roles being clear in the workflow.^[9]
• 录用与时效： 出版方仅公布投稿至首次决定中位 5 天，该口径含编辑部初筛；录用率出版方未公布，库内原记「容易」无官方依据，已删除。Acceptance and speed: the publisher publishes only the median of 5 days from submission to first decision, a figure that includes editorial screening. No acceptance rate is published, and the library's previous note of easy acceptance had no official basis and has been deleted.^[10]
• 学科地位： 该刊被视为机器学习领域历史最久、覆盖面最广的刊物之一，其方法学论文常成为后续算法研究与工程实现的基础引用来源。Standing in the field: the journal is regarded as among the longest-running and broadest-scoped titles in machine learning, its methodological papers routinely serving as base citations for later algorithmic research and engineering implementation.^[11]