BriefGPT.xyz
Oct, 2019
利用局部集成技术检测欠规范性
Detecting Extrapolation with Local Ensembles
HTML
PDF
David Madras, James Atwood, Alex D'Amour
TL;DR
本研究提出了基于局部集成的方法来检测预先训练模型在测试时是否存在欠规范，该方法使用局部二阶信息来计算模型集合的预测方差，并在实验中表现出对检测测试数据中欠规范、检测伪相关性和主动学习具有应用价值的效果。
Abstract
We present
local ensembles
, a method for detecting extrapolation at test time in a
pre-trained model
. We focus on underdetermination as a key component of extrapolation: we aim to detect when many possible predic
→