Industrial Automatic Control Systems and Controllers Annotation << Back Adaptive Stream Processing at the Edge with Active Inference Yarmonov A.S., Gorelikov R.S., Yavtukhovsky E.Yu., Vereshchagina E.A., Frolov A.V. The current scenario of the Internet of Things development shows a constant increase in the volume of data that is generated in a constant stream, which requires new architectural and logical solutions. Accessing the data itself in the computing spectrum guarantees better load distribution and lower latency, as well as increased privacy. Despite the large number of proposals and management decisions based on long-term prediction and control, as well as precise troubleshooting, a new ML paradigm based on active inference (AIF) is proposed for consideration – a neuroscience concept that describes how the brain constantly predicts and evaluates sensory information. Thanks to AIF and its causal structures, the method guarantees full transparency of the decision-making process, which simplifies the interpretation of results and troubleshooting. Keywords: Active inference, machine learning, goal-based serving. DOI: 10.25791/asu.2.2025.1566 Pp. 44-52.