BriefGPT.xyz
Nov, 2017
ConvNets和ImageNet超越准确性：理解错误和揭示偏见
ConvNets and ImageNet Beyond Accuracy: Explanations, Bias Detection, Adversarial Examples and Model Criticism
HTML
PDF
Pierre Stock, Moustapha Cisse
TL;DR
本研究从人类参与和解释性的角度出发，探究了ConvNets和Imagenet在图像分类上的性能、鲁棒性和偏差问题，并以实验和工具提出了解释作为改善模型可靠性和理解性的有效手段。
Abstract
convnets
and
imagenet
have driven the recent success of
deep learning
for image classification. However, the marked slowdown in performanc
→