CGANs
Introducation
1. intruduce the conditional version of GANs, which can be constructed by simply feeding the data , y.
2. the CGANs can be used to learn a multi-modal model.
3.GANs in order to sidestep the difficulty of approximating many intractable probabilistic computations.(为了避免许多难以处理的概率计算的近似困难)
4. Adversarial nets have the advantages that Markov chains are never needed, only backpropagation is used to obtain gradients, no inference is required during learning,
and a wide variety of factors and interactions can easily be incorporated into the model.(多种因素和相互作用可以很容易地纳入模型)
5.马尔可夫链(Markov Chain),描述了一种状态序列,其每个状态值取决于前面有限个状态。一般来说,其核心是满足条件期望和平稳的分布,保证在计算过程中能够得到想要的概率分布。而我们考虑的生成模型恰好可能有以下两种情况:
输入一个随机分布的数据(例如一张黑白像素夹杂的噪音图),输出期望的数据(一张头像)
输入含有噪音的数据(在原有的图像上添加噪点或缺损),输出除去噪点或补完后的数据(完整的原始图像),这种情况下的模型也可以叫做任意去噪的自编码器。
无论是哪种情况,我们都希望从模型输出的数据y的概率分布尽可能逼近训练数据集的概率分布。但是让计算机生成一段音乐,或者一张有意义的图片,这个分布是非常复杂,很难求解的;即使通过马尔可夫链取样,得到了一个生成模型,我们最终也很难对这个模型的效果进行评估,因为生成的音乐到底好不好听,不同的人会得到不同的答案。
6. GANs can produce state of the art log-likehood estimate and realistic samples.
7. but
Related Work
1. the challeage of scaling models to accommodate an extremely large number of predicted output categories (调整模型以适应非常多的预测输出类别的挑战), to adress this problem by leveraging additional information such as using natural language corpora.and even a simple linear mapping from image feature-space to word-representation-space can improve.
2. the challage of focusing on learning one-to-one mapping from input to output,but many interesting problems belong to a probabilistic one-to-many mapping.to adress this challege by using a conditional probabilistic generative model , for example, the input is taken to be the conditioning variable and the one-to-many mapping is instantiated(实例化)as a conditional predictive distribution.
Method
1. to specify that the G can capture the data distribution and the D can estimate the probability that a sample came from the training data rather than G.
2. the input is z, G and D are both trained simultaneously. we adjust the parameters for G to minimize $log(1-D(G(z)))$ and adjust the parameters for D to minimize $log(D(X))$
Import Details -----Conditional Adversarial Nets
The training mechanism of CGANs.
1. GANs can be extended to a conditional model if both the G and D are conditioned on some extra information y.
2. y can be any kind of auxiliary information such as class label or data from other modalities.
3. feeding y into both discriminator and generator as additional input layer.
4. prior input noise and y are combined into joint hidden representation 对抗性训练框架允许在如何组成这种隐藏的表示方面具有相当大的灵活性。
5. In the discriminator and are presented as inputs and to a discriminative function (embodied x y again by a MLP in this case).
The formula of a objective function :
The framework of CGANs:
Experiment
1. this paper trained a CGANs on MNIST images conditioned on their class labels, encoded as one-hot vectors.
For G:
both z and y are mapped to hidden layers with RELU, with layer sizes 200 and 1000 respectively, then combined hidden ReLu layer of dimensionality 1200.
For D:
The discriminator maps to a maxout [6] layer with 240 units and 5 pieces, and to a maxout layer x y with 50 units and 5 pieces. Both of the hidden layers mapped to a joint maxout layer with 240 units and 4 pieces before being fed to the sigmoid layer
For Training:
and best estimate of log-likehood on the validation set was used as stopping point.(并以验证集的对数似然最优估计值作为停止点)。
Summary
CGANs outperforms compared with original GANs, we can combine the class label or data from other modalities into the input of G and D, in order to achieve conditional probabilities distribution and controlling GANs.
CGANs的更多相关文章
- (转)Deep Learning Research Review Week 1: Generative Adversarial Nets
Adit Deshpande CS Undergrad at UCLA ('19) Blog About Resume Deep Learning Research Review Week 1: Ge ...
- Unsupervised Image-to-Image Translation Networks --- Reading Writing
Unsupervised Image-to-Image Translation Networks --- Reading Writing 2017.03.03 Motivations: most ex ...
- Face Aging with Conditional Generative Adversarial Network 论文笔记
Face Aging with Conditional Generative Adversarial Network 论文笔记 2017.02.28 Motivation: 本文是要根据最新的条件产 ...
- #论文笔记# [pix2pixHD] High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs
Ting-Chun Wang, Ming-Yu Liu, Jun-Yan Zhu, Andrew Tao, Jan Kautz, and Bryan Catanzaro. "High-Res ...
- StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation - 1 - 多个域间的图像翻译论文学习
Abstract 最近在两个领域上的图像翻译研究取得了显著的成果.但是在处理多于两个领域的问题上,现存的方法在尺度和鲁棒性上还是有所欠缺,因为需要为每个图像域对单独训练不同的模型.为了解决该问题,我们 ...
- CSAGAN:LinesToFacePhoto: Face Photo Generation from Lines with Conditional Self-Attention Generative Adversarial Network - 1 - 论文学习
ABSTRACT 在本文中,我们探讨了从线条生成逼真的人脸图像的任务.先前的基于条件生成对抗网络(cGANs)的方法已经证明,当条件图像和输出图像共享对齐良好的结构时,它们能够生成视觉上可信的图像.然 ...
- Learning Face Age Progression: A Pyramid Architecture of GANs-1-实现人脸老化
Learning Face Age Progression: A Pyramid Architecture of GANs Abstract 人脸年龄发展有着两个重要的需求,即老化准确性和身份持久性, ...
- AT指令集之Call
1.//unsolicited result code,URC表示BP->AP+ESIPCPI:<call_id>,<dir>,<sip_msg_type>, ...
- mtk 的conferrence call建立流程
(重点看main_log与) 抓mtk log: 1.*#*#82533284#*#* 进入抓log UI 2.*#*#825364#*#* 进入工程模式 3.进入"Lo ...
随机推荐
- stringstream使用
stringstream的头文件是<sstream>,stringstream可以作为中间介质,实现字符串和数字之间的转换. 数字转string double a=213; string ...
- Flutter音频播放--chewie_player的基本使用(二)——样式修改
先贴修改图,只改了部分布局与样式 官方的demo并不十分适合我的需求,从组件进入chewie_player并没有查看到相应的布局,那么直接从chewie的依赖包进入 可以看到以下的目录结构: 我主要修 ...
- Java Web项目获取客户端和服务器的IP地址
在JSP里,获取客户端的IP地址的方法是:request.getRemoteAddr(),这种方法在大部分情况下都是有效的.但是在通过了Apache,Squid等反向代理软件就不能获取到客户端的真实I ...
- 【Processing日常2】群星1
之前在CSDN上发表过: https://blog.csdn.net/fddxsyf123/article/details/79747064
- 《我想进大厂》之MQ夺命连环11问
继之前的mysql夺命连环之后,我发现我这个标题被好多套用的,什么夺命zookeeper,夺命多线程一大堆,这一次,开始面试题系列MQ专题,消息队列作为日常常见的使用中间件,面试也是必问的点之一,一起 ...
- 在KEIL下查看单片机编程内存使用情况
原文链接:https://blog.csdn.net/D_azzle/article/details/83410141 截至到目前为止,本人接触单片机也有将近一年的时间.这一年以来也接触过了很具代表性 ...
- ACMer不得不会的线段树,究竟是种怎样的数据结构?
大家好,欢迎阅读周三算法数据结构专题,今天我们来聊聊一个新的数据结构,叫做线段树. 线段树这个数据结构很多人可能会有点蒙,觉得没有听说过,但是它非常非常有名,尤其是在竞赛圈,可以说是竞赛圈的必备技能. ...
- 080 01 Android 零基础入门 02 Java面向对象 01 Java面向对象基础 01 初识面向对象 05 单一职责原则
080 01 Android 零基础入门 02 Java面向对象 01 Java面向对象基础 01 初识面向对象 05 单一职责原则 本文知识点:单一职责原则 说明:因为时间紧张,本人写博客过程中只是 ...
- sprintf_s() 、sprintf()和printf()区别和用法
转载:https://blog.csdn.net/qq_35608277/article/details/80878802 int sprintf_s(char *buffer,size_t size ...
- Spring Cloud Config配置git私钥出错
重装了电脑之后,重新生成了ssh key文件id_rsa和id_rsa.pub文件. 然后在配置中心的配置了私钥之后启动项目,报错如下: Reason: Property 'spring.cloud. ...