On Learning Prediction-Focused Mixtures

On Learning Prediction-Focused Mixtures
复制标题

关于以预测为中心的混合学习

DOI:
--
复制
发表时间:
2021
期刊:
arXiv.org
影响因子:
--
通讯作者:
F. Doshi
F. Doshi
中科院分区:
--
文献类型:
--
作者:
Abhishek Sharma;Catherine Zeng;Sanjana Narayanan;S. Parbhoo;F. Doshi

文献摘要

参考文献

被引文献

相似文献

概率模型帮助我们对潜在结构进行编码,这些结构既可以对数据进行建模,也可以用于特定的下游任务。其中,混合模型和它们的时间序列对应物,隐马尔可夫模型,识别数据中的离散成分。在这项工作中,我们专注于受限容量设置,我们希望学习具有相对较少组件的模型(例如,出于可解释性目的)。为了保持预测性能,我们引入了以预测为中心的混合建模,它会自动选择与预测任务相关的维度。我们的方法从输入中识别出相关信号,优于不以预测为重点的模型,并且易于优化;我们还描述了以预测为重点的建模何时可以工作。
Probabilistic models help us encode latent structures that both model the data and are ideally also useful for specific downstream tasks. Among these, mixture models and their time-series counterparts, hidden Markov models, identify discrete components in the data. In this work, we focus on a constrained capacity setting, where we want to learn a model with relatively few components (e.g. for interpretability purposes). To maintain prediction performance, we introduce prediction-focused modeling for mixtures, which automatically selects the dimensions relevant to the prediction task. Our approach identifies relevant signal from the input, outperforms models that are not prediction-focused, and is easy to optimize; we also characterize when prediction-focused modeling can be expected to work.
DOI: 10.1001/jama.2019.5791
发表时间: 2019-05-28
影响因子: 120.7
作者:
Seymour, Christopher W.;Kennedy, Jason N.;Angus, Derek C.
通讯作者: Angus, Derek C.
DOI: 10.1016/s2352-3018(20)30340-4
发表时间: 2021-04-01
期刊: LANCET HIV
影响因子: 16.1
作者:
Orkin, Chloe;Oka, Shinichi;Spreen, William R.
通讯作者: Spreen, William R.