发表机构
University of North Carolina at Chapel Hill; Nanyang Technological University; University of Minnesota Twin Cities; University of Michigan（北卡罗来纳大学教堂山分校; 南洋理工大学; 明尼苏达大学双城分校; 密歇根大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对心理学纵向数据预测中参与者负担与预测准确性的矛盾，提出基于树蒸馏的可解释策略，从神经网络LAFA中学习，在模拟和实证数据上以最小精度损失减少获取项目数。
AI 中文摘要
病理结果的准确预测是心理学中的一个核心问题。为此，心理学家经常收集密集的纵向数据。然而，在这类研究中，为了准确预测而获取大量变量的愿望，往往与最小化参与者负担的需求相冲突。每次获取更多变量可以带来更好的预测，但过多的获取会增加无响应和流失的风险。纵向主动特征获取（Longitudinal Active Feature Acquisition, LAFA）是一种解决这一难题的原则性方法。LAFA并非要求在每个获取时机对所有项目作出响应，而是生成一种策略，旨在在每个时间点最优地选择要获取的动态项目子集，同时保持我们预测特定结果的能力。然而，现有的LAFA方法大多基于神经网络（Neural Networks, NN），这些网络在实践中难以解释。在本工作中，我们引入了一种树蒸馏方法，用于从基于NN的LAFA网络中学习可解释的策略。我们通过模拟实验和一个实证的生态瞬时评估（EMA）数据集（用于预测每日酒精消费）验证了我们的方法。在这两种情况下，我们发现可以在准确性损失最小的情况下，显著减少每次获取的项目数量。
英文摘要
Accurate forecasting of pathological outcomes is a central problem in psychology. To do so, psychologists often collect intensive longitudinal data. However, in such studies, the desire to acquire a large number of variables for the sake of accurate prediction is often counteracted by the need to minimize participant burden. Acquiring more variables per occasion can yield better predictions, but having too many acquisitions increase the risk of non-response and attrition. Longitudinal Active Feature Acquisition (LAFA) is a principled approach to resolve this conundrum. Instead of requiring responses to every item at every acquisition occasion, LAFA produces a policy that seeks to optimally select dynamic subsets of items to be acquired at each timepoint while preserving our ability to forecast a specific outcome. However, existing LAFA methods are mostly based on Neural Networks (NN) that are difficult to interpret in practice. In this work, we introduce a tree distillation method for learning an interpretable policy from NN-based LAFA networks. We validated our method through both a simulation and an empirical EMA dataset on forecasting daily alcohol consumption. In both cases, we find that we can meaningfully reduce the number of items acquired at each occasion with minimal loss in accuracy. Networks (NN) that are difficult to interpret in practice. In this work, we introduce a tree distillation method for learning an interpretable policy from NN-based LAFA networks. We validated our method through both a simulation and an empirical EMA dataset on forecasting daily alcohol consumption. In both cases, we find that we can meaningfully reduce the number of items acquired at each occasion with minimal loss in accuracy.
