<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[pandav5的专栏]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/pandav5</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; pandav5]]></copyright><item><title><![CDATA[sci-ei求中啊啊啊]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/77998575</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/77998575</guid><author>pandav5</author><pubDate>Sat, 16 Sep 2017 08:57:53 +0800</pubDate><description><![CDATA[pattern analysis and machine intelligence （PAMI） 顶级
International Journal of Computer Vision  （IJCV） 顶级
SIAM Journal on Imaging Sciences 顶级
Pattern Recognition (PR) 2.509 高
IEEE TRANSACTIONS ON IM]]></description><category></category></item><item><title><![CDATA[Active learning-Fine-tuning Convolutional Neural Networks for Biome﻿dical Image Analysis: Actively a]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/72594192</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/72594192</guid><author>pandav5</author><pubDate>Sun, 21 May 2017 10:18:50 +0800</pubDate><description><![CDATA[转 ：  http://zongweiz.leanote.com/post/Active-Learning-%E4%B8%80%E4%B8%AA


作者 ：周纵苇


﻿下面要介绍的工作发表于CVPR2017，题为﻿﻿“Fine-tuning
 Convolutional Neural Networks for Biome﻿dical Image Analysis: Activel]]></description><category></category></item><item><title><![CDATA[交叉熵与相对熵 ---转载]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/72236863</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/72236863</guid><author>pandav5</author><pubDate>Tue, 16 May 2017 11:14:32 +0800</pubDate><description><![CDATA[选至 知乎 ：https://www.zhihu.com/question/41252833]]></description><category></category></item><item><title><![CDATA[牛人的博客（图像处理，机器视觉，机器学习等）]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/55212394</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/55212394</guid><author>pandav5</author><pubDate>Wed, 15 Feb 2017 17:16:21 +0800</pubDate><description><![CDATA[1、小魏的修行路  http://blog.youkuaiyun.com/xiaowei_cqu  

2、晨宇思远（邹宇华 北京航空航天大学） http://blog.youkuaiyun.com/chenyusiyuan 

3、Rachel Zhang（张睿卿）的博客 http://blog.youkuaiyun.com/abcjennifer 

4、zouxy09（邹晓艺） 华南理工大学 http://blog]]></description><category></category></item><item><title><![CDATA[Loss Function view---关于损失函数的介绍]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/54344154</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/54344154</guid><author>pandav5</author><pubDate>Wed, 11 Jan 2017 15:42:06 +0800</pubDate><description><![CDATA[一、Loss Function

什么是Loss Function？wiki上有一句解释我觉得很到位，引用一下：The loss function quantifies the amount by which the prediction deviates from the actual values。Loss Function中文损失函数，适用于用于统计，经济，机器学习等领域，虽外表形式不一]]></description><category></category></item><item><title><![CDATA[关于Mxnet的一些基础知识理解(1)]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/53993684</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/53993684</guid><author>pandav5</author><pubDate>Tue, 03 Jan 2017 16:54:11 +0800</pubDate><description><![CDATA[(1)Mxnet的数据格式为NDArray，当需要读取可观看的数据，就要调用：

numpy_d = d.asnumpy()

converts
 it to a Numpy array.

(2)list_arguments （给出当前符号d的输入变量）与list_outputs(给出符号d的输出变量)的说明

import mxnet as mx
a = mx.sym.Var]]></description><category></category></item><item><title><![CDATA[detection总结R-CNN>SPP-Net>Fast-R-CNN>Faster-R-CNN>YOLO>SSD]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/53671647</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/53671647</guid><author>pandav5</author><pubDate>Thu, 15 Dec 2016 16:04:54 +0800</pubDate><description><![CDATA[1. R-CNN：Rich feature hierarchies for accurate object detection and semantic segmentation



论文：https://arxiv.org/abs/1311.2524

代码：https://github.com/rbgirshick/rcnn
技术路线：selective search + CN]]></description><category></category></item><item><title><![CDATA[如何做好计算机视觉的研究]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/53537711</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/53537711</guid><author>pandav5</author><pubDate>Fri, 09 Dec 2016 13:24:35 +0800</pubDate><description><![CDATA[要回答这个问题，我们先要对这个问题的关键词进行分析：如果去掉“计算机视觉”这个限定词，这问题就变成了“如何做好研究？”那么，要回答这个问题，我们就要知道“什么是好的研究？”而要定义什么是好的研究，必须回到根本，先要知道“什么是研究？”


先确定狭义的研究的定义为：利用科学的方法来调查解释一个现象或者获取新的知识。



科学研究从本质上是由三个基本的要素构成：
1) 目的：产生新]]></description><category></category></item><item><title><![CDATA[Q-learning--定义--过程]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/53099197</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/53099197</guid><author>pandav5</author><pubDate>Wed, 09 Nov 2016 14:16:11 +0800</pubDate><description><![CDATA[The Q-Learning algorithm was proposed as a way to optimize solutions in Markov decision process problems.  The distinctive feature of Q-Learning is in its capacity to choose between immediate rewards ]]></description><category></category></item><item><title><![CDATA[机器学习需看书籍]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/53098720</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/53098720</guid><author>pandav5</author><pubDate>Wed, 09 Nov 2016 13:18:21 +0800</pubDate><description><![CDATA[如果你只愿意看一本书，那么推荐Bishop的PRML，全名Pattern Recognition and Machine Learning. 这本书是机器学习的圣经之作，尤其对于贝叶斯方法，介绍非常完善。该书也是众多高校机器学习研究生课程的教科书，如Deep Learning的开山鼻祖Hilton在多伦多大学教授机器学习课程（CSC2515）时也选择了这本书。该书电子版在Bishop个人网页提供直]]></description><category></category></item><item><title><![CDATA[mxnet学习记录（转载）]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/52766550</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/52766550</guid><author>pandav5</author><pubDate>Sun, 09 Oct 2016 14:33:39 +0800</pubDate><description><![CDATA[转载来至
http://blog.youkuaiyun.com/shuzfan/article/details/50037273



由于caffe依赖性太多，配置极其复杂，所以将接下来的学习转向mxnet.因此本文主要记录我的学习历程，如果描述有什么问题，欢迎大家的指正。

mxnet的优点很明显，简洁灵活效率高 ，多机多卡支持好。

mxnet的github下载链接：https://g]]></description><category></category></item><item><title><![CDATA[在VS2010下编译libjpeg及其应用例子]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51645147</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51645147</guid><author>pandav5</author><pubDate>Sun, 12 Jun 2016 12:00:26 +0800</pubDate><description><![CDATA[原地址来之：

http://www.360doc.com/content/13/1122/18/13084517_331351714.shtml





下载到libjpeg源码后，解压，可以看到很多文档，想在VS2010中编辑使用，主要参考文件：install.txt
在字段Microsoft Windows, Microsoft Visual C++ 2010 Devel]]></description><category></category></item><item><title><![CDATA[Deep Neural Networks的Tricks~~翻译版~~精华]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51178032</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51178032</guid><author>pandav5</author><pubDate>Tue, 19 Apr 2016 08:51:33 +0800</pubDate><description><![CDATA[Here we will introduce these extensive implementation details, i.e., tricks or tips, for building and training your own deep networks.


主要以下面八个部分展开介绍：
mainly in eight
 aspects: 1) data
 augment]]></description><category></category></item><item><title><![CDATA[Galaxy ZOO银河星空图的识别分类]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51158662</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51158662</guid><author>pandav5</author><pubDate>Fri, 15 Apr 2016 10:28:48 +0800</pubDate><description><![CDATA[本文是以2015年，Galaxy-zoo星空图分类展开叙述，主要归纳了论文中的重点方法和技巧
http://benanne.github.io/2014/04/05/galaxy-zoo.html



挑战赛： Kaggle 
素材来源：Galaxy
 Zoo users (zooites) would classify images of galaxies from the Sl]]></description><category></category></item><item><title><![CDATA[自己整理的计算机视觉领域稍微容易中的期刊（第一版）]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51121828</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51121828</guid><author>pandav5</author><pubDate>Mon, 11 Apr 2016 15:06:33 +0800</pubDate><description><![CDATA[http://blog.youkuaiyun.com/carson2005/article/details/6292748





模式识别，计算机视觉领域，期刊

 

(1)pattern recognition letters, 从投稿到发表，一年半时间

(2)Pattern recognition 不好中，时间长

(3)IEICE Transactions on In]]></description><category></category></item><item><title><![CDATA[发表SCI或EI类英文文章的一些投稿经验]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51121369</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51121369</guid><author>pandav5</author><pubDate>Mon, 11 Apr 2016 14:19:09 +0800</pubDate><description><![CDATA[英文投稿的一点经验【转载】

1. 首先一定要注意杂志的发表范围, 超出范围的千万别投,要不就是浪费时间;另外，每个杂志都有他们的具体格式要求，一定要按照他们的要求把论文写好，免得浪费时间，前些时候，我的一个同事向一个著名的英文杂志投稿，由于格式问题，人家过两个星期就退回来了，而且说了很多难听的话，说投稿前首先就应该看清楚他们的格式要求；
2. 论文写作一定要言简意赅,特别是摘要,引言和结论]]></description><category></category></item><item><title><![CDATA[回归总结，logistic,一般线性，softmax回归等]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51119131</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51119131</guid><author>pandav5</author><pubDate>Mon, 11 Apr 2016 09:41:31 +0800</pubDate><description><![CDATA[http://blog.youkuaiyun.com/statdm/article/details/7585153


本博文由  JerryLead 所写，对初学者有所帮组，故引入。





1 摘要

      本报告是在学习斯坦福大学机器学习课程前四节加上配套的讲义后的总结与认识。前四节主要讲述了回归问题，回归属于有监督学习中的一种方法。该方法的核心思想是从连续型统计数据中得到]]></description><category></category></item><item><title><![CDATA[总结径向基函数，径向基网络，计算过程，训练过程介绍（参考+翻译）]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51103196</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51103196</guid><author>pandav5</author><pubDate>Sat, 09 Apr 2016 10:45:54 +0800</pubDate><description><![CDATA[1、SVM就是个MLP（多层神经网络）


2、RBF径向基核函数在SVM发明的7年前就已经被用于RBF神经网络，RBF网络本质就是个把激活函数从Sigmoid替换成RBF的MLP。
RBF的结构：
径向基函数（Radical
 Basis Function，RBF）方法是Powell在1985年提出的。

详细看这篇：
http://blog.youkuaiyun.com/zouxy09/]]></description><category></category></item><item><title><![CDATA[各种优化方法总结比较（sgd/momentum/Nesterov/adagrad/adadelta）]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51058402</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51058402</guid><author>pandav5</author><pubDate>Mon, 04 Apr 2016 20:45:37 +0800</pubDate><description><![CDATA[转载地址：
http://blog.youkuaiyun.com/luo123n/article/details/48239963


http://sebastianruder.com/optimizing-gradient-descent/







这里讨论的优化问题指的是，给定目标函数f(x)，我们需要找到一组参数x，使得f(x)的值最小。

本文以下内容假设读者已经了]]></description><category></category></item><item><title><![CDATA[Jackknife，Bootstraping, bagging, boosting, AdaBoosting, Rand forest 和 gradient boosting的简单介绍]]></title><link>https://blog.youkuaiyun.com/pandav5/article/details/51004840</link><guid>https://blog.youkuaiyun.com/pandav5/article/details/51004840</guid><author>pandav5</author><pubDate>Tue, 29 Mar 2016 12:34:35 +0800</pubDate><description><![CDATA[转之

 http://blog.youkuaiyun.com/chenhongc/article/details/9404583

http://blog.sina.com.cn/s/blog_5dd2e9270100c8ko.html





在这里致谢作者的研究。

一并列出一些找到的介绍boosting算法的资源：

（1）视频讲义，介绍boosting算法，主要介绍Ad]]></description><category></category></item></channel></rss>