Decentralized machine learning is a research area and methodological paradigm concerned with training machine learning models across multiple distributed entities or devices without reliance on a single, central server for data aggregation or model coordination. This field investigates algorithms, architectures, and protocols that enable collaborative model development while participants retain local control over their data, addressing challenges related to privacy, security, communication efficiency, and robustness in distributed network topologies. Its significance lies in facilitating privacy-preserving collaborative AI and mitigating single points of failure inherent in centralized systems.