This website requires JavaScript.

Online Active Learning for Soft Sensor Development using Semi-Supervised Autoencoders

Davide CacciarelliMurat KulahciJohn Tyssedal
Dec 2022
摘要
Data-driven soft sensors are extensively used in industrial and chemicalprocesses to predict hard-to-measure process variables whose real value isdifficult to track during routine operations. The regression models used bythese sensors often require a large number of labeled examples, yet obtainingthe label information can be very expensive given the high time and costrequired by quality inspections. In this context, active learning methods canbe highly beneficial as they can suggest the most informative labels to query.However, most of the active learning strategies proposed for regression focuson the offline setting. In this work, we adapt some of these approaches to thestream-based scenario and show how they can be used to select the mostinformative data points. We also demonstrate how to use a semi-supervisedarchitecture based on orthogonal autoencoders to learn salient features in alower dimensional space. The Tennessee Eastman Process is used to compare thepredictive performance of the proposed approaches.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答