EVENT DOSSIER
To address the hidden confounding issues caused by missing variables in wireless network observations, researchers proposed a new method called Confounding-Valid Counterfactual Conformal Inference (CV-CCI). This method combines the principles of General Synthetic-Powered Inference (GESPI) and integrates rich potential confounding observational data with limited randomized data. Experiments show that CV-CCI remains effective in the presence of hidden confounding in two typical wireless access network control tasks and generates a prediction set more efficient than existing baseline methods.
研究人员提出了一种名为 Confounding-Valid Counterfactual Conformal Inference（CV-CCI）的方法，用于解决无线网络中因观测数据缺失变量导致的隐藏混杂问题。该方法结合 General Synthetic-Powered Inference（GESPI）原则，将丰富的潜在混杂观测遥测数据与有限的随机化数据进行融合。实验表明，CV-CCI 在两个典型的无线接入网（RAN）控制任务中，能在存在隐藏混杂的情况下保持有效性，并生成比现有基线方法效率更高的预测集。