BriefGPT.xyz
Nov, 2017
深度视觉解释简介
An Introduction to Deep Visual Explanation
HTML
PDF
Housam Khalifa Bashier Babiker, Randy Goebel
TL;DR
研究了深度学习方法面临的可解释性困境,提出了一种叫做“深度视觉解释”的框架,通过可视化技术来识别和暴露深度学习模型性能行为的假设,以提高模型的可解释性和可调试性,并展示了初步的可解释性应用实验结果。
Abstract
The practical impact of
deep learning
on complex supervised learning problems has been significant, so much so that almost every Artificial Intelligence problem, or at least a portion thereof, has been somehow recast as a
→