[论文笔记]Cognitive Psychology for Deep Neural Networks: A Shape Bias Case Study
研究背景 神经网络的可解释性逐渐成为研究热点，本文的研究人员认为神经网络决策过程与人类认知过程也许有相似之处，并利用已经提出的人类认知过程假设来研究神经网络的决策过程。 主要贡献 本文从量化角度证明了神经网络识别时更偏好于识别形状而非颜色。 证明这种偏差在同结构但不同初始化的网络中存在极大差异，并且和分类正确率无关。引申结论即为不同初始化的网络具有不同的决策过程，即使分类正确率类似。（笔者认为类似
相关文章
- 1. Understanding Neural Networks Through Deep Visualization 论文笔记
- 2. 【论文笔记】A Survey on Deep Learning for Named Entity Recognition
- 3. [论文解读] Concolic Testing for Deep Neural Networks
- 4. 论文笔记：Quantizing deep convolutional networks for efficient inference: A whitepaper
- 5. Deep Neural Networks for Learning Graph Representations论文笔记
- 6. Learning Convolutional Neural Networks for Graphs论文笔记
- 7. 论文笔记：Filter Pruning via Geometric Median for Deep Convolutional Neural Networks Acceleration
- 8. 论文笔记——Aggregated Residual Transformations for Deep Neural Networks(ResNeXt)
- 9. Deep Neural Networks for Object Detection论文翻译
- 10. 论文笔记：Learning a Wavelet-like Auto-Encoder to Accelerate Deep Neural Networks
- 更多相关文章...
- • Scala for循环 - Scala教程
- • ASP.NET Razor - 标记 - ASP.NET 教程
- • Tomcat学习笔记（史上最全tomcat学习笔记）
- • Scala 中文乱码解决
相关标签/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
最新文章
欢迎关注本站公众号,获取更多信息
相关文章
- 1. Understanding Neural Networks Through Deep Visualization 论文笔记
- 2. 【论文笔记】A Survey on Deep Learning for Named Entity Recognition
- 3. [论文解读] Concolic Testing for Deep Neural Networks
- 4. 论文笔记：Quantizing deep convolutional networks for efficient inference: A whitepaper
- 5. Deep Neural Networks for Learning Graph Representations论文笔记
- 6. Learning Convolutional Neural Networks for Graphs论文笔记
- 7. 论文笔记：Filter Pruning via Geometric Median for Deep Convolutional Neural Networks Acceleration
- 8. 论文笔记——Aggregated Residual Transformations for Deep Neural Networks(ResNeXt)
- 9. Deep Neural Networks for Object Detection论文翻译
- 10. 论文笔记：Learning a Wavelet-like Auto-Encoder to Accelerate Deep Neural Networks