AI & ML
The Prediction-to-Decision Gap: First-Action Informativity in Data-Driven Predictive Control
arXiv cs.SY0 views
arXiv:2611.25319v1 Announce Type: new Abstract: Model predictive control (MPC) applies only the first input of an optimized sequence before replanning. Do the available data determine this input? We define first-action informativity as agreement across data-compatible models at a fixed history and reference, with common MPC settings. For autoregressive models with exogenous inputs (ARX) and unconstrained quadratic MPC, we show that first-action agreement and prediction agreement under a fixed input sequence do not imply one another. We give structural conditions for action agreement without requiring unique model identification. When actions disagree, we quantify their mean-squared deviation from the fitted model's input under a specific parameter distribution. We approximate this measure using local sensitivities and parameter covariance. Analytical and numerical examples illustrate how model uncertainty affects predictions and actions differently. The framework evaluates model uncertainty through its effect on the first action MPC is about to apply.
Source:https://arxiv.org/abs/2611.25319PreviousData-driven Linearization of Operational Dynamics of Soft Open Points in Power Distribution NetworksNextTrace2ST: Recovering PLC Structured Text from I/O Traces with LLM Generation and Verification
Related posts

Serving Qwen3.5-0.8B with a 0.016% weight edit: measure the loops, pin the sampling on the server
0dev.to
Design Tokens for AI: Extract a Website's Real Colours, Fonts, Shadows and Icons So an LLM Builds On Brand
0dev.to
Trying to get our Arms around AI efficiency
0Stack Overflow Blog
Why an Embedded Game Fails on Some Phones, and How to Debug It
0dev.to
Microsoft 365 vs Office 2024: The Break-Even Math
0dev.to
Master AI Evals: Build Reliable, Scalable Model Evaluation Pipelines
0dev.to