Reading Note: Interpretable Convolutional Neural Networks
来源:互联网 发布:网络电视客户端 编辑:程序博客网 时间:2024/06/04 23:28
TITLE: Interpretable Convolutional Neural Networks
AUTHOR: Quanshi Zhang, Ying Nian Wu, Song-Chun Zhu
ASSOCIATION: UCLA
FROM: arXiv:1710.00935
CONTRIBUTION
- Slightly revised CNNs are propsed to improve their interpretability, which can be broadly applied to CNNs with different network structures.
- No annotations of object parts and/or textures are needed to ensure each high-layer filter to have a certain semantic meaning. Each filter automatically learns a meaningful object-part representation without any additional human supervision.
- When a traditional CNN is modified to an interpretable CNN, experimental settings need not to be changed for learning. I.e. the interpretable CNN does not change the previous loss function on the top layer and uses exactly the same training samples.
- The design for interpretability may decrease the discriminative power of the network a bit, but such a decrease is limited within a small range.
METHOD
The loss for filter is illustrated in the following figure.
{: .center-image .image-width-480}
A feature map is expected to be strongly activated in images of a certain category and keep silent on other images. Therefore, a number of templates are used to evaluate the fitness between the current feature map and the ideal distribution of activations w.r.t. its semantics. The template is an ideal distribution of activations according to space locations. The loss for layers is formulated as the mutual information between feature map
the loss can be re-written as
The first term is a constant denoting the piror entropy of
SOME THOUGHTS
This loss can reduce the redundancy among filters, which may be used to compress the model.
- Reading Note: Interpretable Convolutional Neural Networks
- Interpretable Convolutional Neural Networks笔记
- READING NOTE: Factorized Convolutional Neural Networks
- READING NOTE: Chained Predictions Using Convolutional Neural Networks
- READING NOTE: Do semantic parts emerge in Convolutional Neural Networks?
- READING NOTE: Object Detection from Video Tubelets with Convolutional Neural Networks
- READING NOTE: Spatially Supervised Recurrent Convolutional Neural Networks for Visual Object Trackin
- Reading Note: MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
- READING NOTE: Densely Connected Convolutional Networks
- READING NOTE: Weakly Supervised Cascaded Convolutional Networks
- Convolutional Neural Networks Learning Note (一)
- READING NOTE:LCNN: Lookup-based Convolutional Neural Network
- READING NOTE: Aggregated Residual Transformations for Deep Neural Networks
- Convolutional Neural Networks
- Convolutional Neural Networks (LeNet)
- Convolutional Neural Networks Posts
- Convolutional neural networks
- Convolutional Neural Networks学习
- tcp-三次握手与四次挥手
- ofbiz后台系统二级菜单的配置——组件菜单
- VC++ 比较字符串是否相等的函数strcmp和wcscmp
- 【30】纯CSS制作一台打字的电脑
- Git -- cherry-pick的用法
- Reading Note: Interpretable Convolutional Neural Networks
- recyclerView三种管理器和删除添加功能
- css3属性-webkit-font-smoothing
- kylin restfulApi
- ios常用三方的podfile
- 抽屉布局(侧拉)
- git忽略已经被提交的文件
- 初学Java,用notepad++时遇到的编码问题
- Spring 4.2集成 Quartz Scheduler 2.2.2 任务调度简单实现示例