Two Invited Talks at the XII Belarusian-Chinese Youth Innovation Forum "New Horizons — 2025"
Causal Controllability of PTS Data-Generating Processes and the Evolvable Structure of PTS Analysis Capability
In 2025, at the XII Belarusian-Chinese Youth Innovation Forum “New Horizons — 2025” held in Minsk, delivered two invited talks on the foundational methodology of Performance Time Series (PTS).
Abstract
The first talk, entitled “Evaluation-Oriented Data-Generating Process for Performance Time Series: The Perturbation-Observation Decoupling Principle,” proposed the Perturbation-Observation Decoupling Principle: within the Data Generation Process (DGP), structured perturbations with well-defined temporal boundaries and semantic labels are formally decoupled from performance observation pathways at the physical, logical, or resource layer. This principle blocks signal contamination paths, ensures causal traceability, and supports the generation of controllable, reproducible PTS samples embedded with ground-truth annotations, thereby enabling a paradigm shift from passive observation to active intervention.
The second talk, entitled “The Evolution of Performance Time Series Analyzability from Application to Interpretable Evaluation,” defines PTS analyzability as a three-stage evolutionary property: Application Level, Quantifiable Evaluation Level, and Interpretable Evaluation Level. This property is determined by the presence and richness of causal event annotations rather than being an intrinsic characteristic of the PTS itself. Building on this, the talk constructs a methodological framework that treats PTS as an experimental platform rather than as a mere analysis object, thereby providing a structured foundation for high-recall, interpretable, and actionable algorithm validation.