Skip to content
    Flow-alerta.cab

    Federated Learning

    AI fundamentals & modelsdistributed training without central raw data
    Editorially reviewed on: 2 sources
    Your reading view

    The short answer · 15 words

    In short

    Federated Learning trains a common model across distributed data sources without centrally merging raw data.

    Flow-alerta.cab (2026): Federated Learning. https://flow-alerta.cab/en/glossar/federated-learning. Editorially reviewed on: 10 October 2026.

    How much detail would you like?

    Practical

    It can connect data silos, but requires protection against leakage, manipulation and heterogeneous data.

    Hypothetical example: Several works train a fault model locally and share safe model updates.

    In the case of cross-location SME pilots, you should jointly check governance, data protection, appropriate protection procedures and communication effort.

    At a glance

    Core
    Participants calculate local updates; often a central coordinator aggregates them into a common model. This avoids the central collection of raw data, but alone does not guarantee privacy or security.
    Use
    It can connect data silos, but requires protection against leakage, manipulation and heterogeneous data.
    Example
    Hypothetical example: Several works train a fault model locally and share safe model updates.
    Limits
    Adopt privacy automatically; unsafe updates; unequal data distributions; ignore communication costs and participant failures.

    In context

    Edge AI

    Execution of AI algorithms directly on local devices or sensors instead of in central cloud data centers.

    Questions about this term

    Does data always remain private?

    Raw data remains local, but updates can still reveal information.

    Do you need a server?

    Often an aggregator.

    All details and background

    Participants calculate local updates; often a central coordinator aggregates them into a common model. This avoids the central collection of raw data, but alone does not guarantee privacy or security.

    ELI5 – Explained simply

    Many locations learn together while their raw data remains on site.

    Why it matters in practice

    It can connect data silos, but requires protection against leakage, manipulation and heterogeneous data.

    Example

    Hypothetical example: Several works train a fault model locally and share safe model updates.

    Pitfalls and limits

    • Adopt privacy automatically; unsafe updates; unequal data distributions; ignore communication costs and participant failures.

    Origin & history

    The term was coined by Google in 2016 for decentralized learning on devices; related distributed procedures are older.

    Use in mid-sized companies

    • In the case of cross-location SME pilots, you should jointly check governance, data protection, appropriate protection procedures and communication effort.

    Related terms

    • Secure aggregation
    • Differential privacy
    • Distributed learning
    • Data sovereignty
    • Edge AI

    Matching Flow-alerta service

    AI opportunity analysis & roadmapAI consulting: Examine the prerequisites and limits of a cross-location learning project.

    Matching tool / use case

    View use casesUse Case: Cross-site pilot with safe aggregation

    Sources

    Quellenprüfung: 10.10.2026. Definition und dokumentierte Einordnung abgeglichen; Beispiele beschreiben mögliche Anwendungen.

    Editorially reviewed on:

    Next step

    Put knowledge into practice

    Discuss your projectView use cases