<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[燃烧吧！我的小宇宙！]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/foolsnowman</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; foolsnowman]]></copyright><item><title><![CDATA[快捷键]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/78763001</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/78763001</guid><author>foolsnowman</author><pubDate>Sun, 10 Dec 2017 08:22:06 +0800</pubDate><description><![CDATA[chrome浏览器com + t 打开新的标签页 
com + w 关闭当前标签页shellctrl + u  删除当前命令 
ctrl + a （ahead）移动到行首 
ctrl + e系统com + space 打开搜索 
com + opt + esc 打开强制退出应用程序]]></description><category></category></item><item><title><![CDATA[python中软件包安装]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/55656840</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/55656840</guid><author>foolsnowman</author><pubDate>Sat, 18 Feb 2017 10:23:41 +0800</pubDate><description><![CDATA[Python搜索路径python -c 'import os;print os.sys.path'输出['', '/usr/lib/python2.7', '/usr/lib/python2.7/plat-x86_64-linux-gnu', '/usr/lib/python2.7/lib-tk', '/usr/lib/python2.7/lib-old', '/usr/lib/python2.7/]]></description><category></category></item><item><title><![CDATA[编程实现]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/53817319</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/53817319</guid><author>foolsnowman</author><pubDate>Thu, 22 Dec 2016 17:15:33 +0800</pubDate><description><![CDATA[memoryout-of-memory之类的错误需要看看是否有一些被ctrl+z的任务仍然活着，这时候要kill -9实现算法需要考虑的问题1逻辑上封装 
2对库的功能准确地把握，然后调用库]]></description><category></category></item><item><title><![CDATA[SciPy]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/53815715</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/53815715</guid><author>foolsnowman</author><pubDate>Thu, 22 Dec 2016 15:51:55 +0800</pubDate><description><![CDATA[SciPy在NumPy基础上构建的用于科学计算的Python库。稀疏矩阵导入包from scipy.sparse import csr_matrixcsr存储开销 2*num_nuz+num_col+1 ，csc存储开销 2*num_nuz+num_row+1 所以对于shape(big_row,1)稀疏矩阵应该采用csr的存储方式。]]></description><category></category></item><item><title><![CDATA[caffe之 Blob、Layer、Net]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/53326935</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/53326935</guid><author>foolsnowman</author><pubDate>Thu, 24 Nov 2016 22:12:28 +0800</pubDate><description><![CDATA[Blobshare_ptr<Blob<float> > b;
Blob<float> bf=*b;//Error,'='操作是Blob私有的LayerNet首先从初始化一个netparameter对象说起，对应caffe.proto中的message为NetParameter。]]></description><category></category></item><item><title><![CDATA[Matlab入门]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/53012345</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/53012345</guid><author>foolsnowman</author><pubDate>Wed, 02 Nov 2016 20:50:01 +0800</pubDate><description><![CDATA[转自：http://blog.youkuaiyun.com/predict_wise/article/details/49077939 
下载matlab安装包 
下载安装包以及Crack 
Crack中包含license和破解文件 
解压得到iso镜像,并挂载到linux$sudo mkdir /download/temp_matlab (此文件夹用于临时存放挂载后的iso文件)
$sudo mount -o]]></description><category></category></item><item><title><![CDATA[docker]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52566742</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52566742</guid><author>foolsnowman</author><pubDate>Sat, 17 Sep 2016 23:07:19 +0800</pubDate><description><![CDATA[文件拷贝docker run -it -v /home/user/trans:/opt/trans face]]></description><category></category></item><item><title><![CDATA[window<->linux移植]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52529235</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52529235</guid><author>foolsnowman</author><pubDate>Tue, 13 Sep 2016 20:33:55 +0800</pubDate><description><![CDATA[itoa原型：char  *itoa(int   value,char   *string,int   radix) 
用法：#include     功能:将整数value转换成字符串存入string, radix为转换时所用基数(保存到字符串中的数据的进制基数 2 8 10 16)说明：返回指向转换后的字符串的指针 
如：int  number   =   12345;    
char str]]></description><category></category></item><item><title><![CDATA[opencv]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52503850</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52503850</guid><author>foolsnowman</author><pubDate>Sun, 11 Sep 2016 14:35:14 +0800</pubDate><description><![CDATA[First exampleC++#include<opencv2/highgui/highgui.hpp>
using namespace cv;int main()
{
    Mat img = imread("/home/USER/Pictures/python.jpg",CV_LOAD_IMAGE_COLOR);
    imshow("opencvtest",img);
    waitK]]></description><category></category></item><item><title><![CDATA[caffe 预测]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52494419</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52494419</guid><author>foolsnowman</author><pubDate>Sat, 10 Sep 2016 12:55:51 +0800</pubDate><description><![CDATA[使用caffemodel做预测C++ APIsudo ../../build/examples/cpp_classification/classification.bin  VGG_FACE_deploy.prototxt VGG_FACE.caffemodel ../../data/ilsvrc12/imagenet_mean.binaryproto names.txt ak.png详细见官网]]></description><category></category></item><item><title><![CDATA[算法题]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52463439</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52463439</guid><author>foolsnowman</author><pubDate>Wed, 07 Sep 2016 21:25:56 +0800</pubDate><description><![CDATA[1.贪心算法只考虑当前最优解，即每次求解局部最优解。可以用深度优先搜索来类比贪心算法的过程，只是深搜，所以可能得到的不是全局最优解。而要想得到全局优，求解的问题必须具有无后效性，也就是说，以后的过程不会影响之前的状态，用局部最优解来合成全局最优。该类算法的复杂度与子问题的个数，即树深，有关。 
比如, 
两个最小生成树算法：prim、kruskal,Dijkstra（最短路径算法）2.动态规划将原问]]></description><category></category></item><item><title><![CDATA[特征选择]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52431491</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52431491</guid><author>foolsnowman</author><pubDate>Sun, 04 Sep 2016 11:46:35 +0800</pubDate><description><![CDATA[从特征选择和随后的模型训练的关系上，主要分为三大类方法:过滤式、包裹式、嵌入式[1].1. 过滤式特征子集搜索与评价与模型训练过程不重合，将过滤得到的特征用于训练。其中搜索特征子集合的方法主要有：前向、后向、双向；模型评价一般基于信息熵理论。2. 包裹式通过机器学习算法本身对模型进行评价。比如，LVW(Las Vegas Wrapper)拉斯维加斯特征选择方法。3. 嵌入式基于L1、L2正则化的机器]]></description><category></category></item><item><title><![CDATA[java容器类小结]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52268115</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52268115</guid><author>foolsnowman</author><pubDate>Sun, 21 Aug 2016 16:53:47 +0800</pubDate><description><![CDATA[java容器类可以持有对象、并且容量大小可以在运行时才决定，而数组的大小是在编译时就要固定下来的。所以java容器类在容量大小上具有较高的灵活性。包含两类Collection、Map.1.Collection1.1QueuePriorityQueue实际是一个小根堆。对于n个结点的无序序列建堆的时间复杂度为O(n),之后输出正确顺序时需要(n-1)次向下调整，每次调整的时间复杂度是O(logn),故]]></description><category></category></item><item><title><![CDATA[神经网络激活函数与损失函数]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52181210</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52181210</guid><author>foolsnowman</author><pubDate>Thu, 11 Aug 2016 11:42:23 +0800</pubDate><description><![CDATA[sigmoid输出层使用的损失函数为cross-entropysoftmax 输出层使用的损失函数为log-likelihood]]></description><category></category></item><item><title><![CDATA[机器学习与统计学]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52062740</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52062740</guid><author>foolsnowman</author><pubDate>Fri, 29 Jul 2016 11:04:41 +0800</pubDate><description><![CDATA[术语对比
  Machine Learning
  Statistics  network, graphs
  model
  weights
  parameters
  learning
  fitting
  supervised learning
  regression/classiﬁcation
  unsupervised learning
  density estimation,]]></description><category></category></item><item><title><![CDATA[分类]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/52052501</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/52052501</guid><author>foolsnowman</author><pubDate>Thu, 28 Jul 2016 10:14:00 +0800</pubDate><description><![CDATA[广义线性模型一般线性回归logistic回归对p(Y|X)p(Y|X)和XX之间的关系建模。朴素贝叶斯分类朴素贝叶斯分类的主要思想是：对于待分类数据X=(x1,x2,...,xm)X=(x_1,x_2,...,x_m)，计算使后验概率p(Y=cj|X)p(Y=c_j|X)最大的YY的取值，即为数据XX所属的类别标签cjc_j. 
后验概率p(Y=cj|X)p(Y=c_j|X)是通过贝叶斯公式计算的，]]></description><category></category></item><item><title><![CDATA[参数估计方法]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/51984055</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/51984055</guid><author>foolsnowman</author><pubDate>Thu, 21 Jul 2016 16:24:15 +0800</pubDate><description><![CDATA[极大似然估计（Maximum likelihood estimation,ML）
最大后验概率(Maximum A Posteriori Probability Estimation,MAP)估计
贝叶斯估计(Bayesian Estimation)]]></description><category></category></item><item><title><![CDATA[SVM(1)-概念与理解]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/51921342</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/51921342</guid><author>foolsnowman</author><pubDate>Fri, 15 Jul 2016 22:06:51 +0800</pubDate><description><![CDATA[发展理解低纬度空间转到高维空间]]></description><category></category></item><item><title><![CDATA[VC理论]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/51908905</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/51908905</guid><author>foolsnowman</author><pubDate>Thu, 14 Jul 2016 14:14:37 +0800</pubDate><description><![CDATA[VC 理论是Vladimir Vapnik（俄）和Alexey Chervonenkis在1960-1990逐步发展起来的，属于计算学习理论的一种，从概率统计的角度解释学习过程。 
关于该理论最早的发表是Vapnik 和Chervonenkis, 1968，最早的证明是 Vapnik and Chervonenkis, 1971. 
被Sauer，于1972年进一步证明。参考文献Vapnik，Che]]></description><category></category></item><item><title><![CDATA[R作图]]></title><link>https://blog.youkuaiyun.com/foolsnowman/article/details/51889917</link><guid>https://blog.youkuaiyun.com/foolsnowman/article/details/51889917</guid><author>foolsnowman</author><pubDate>Tue, 12 Jul 2016 16:06:18 +0800</pubDate><description><![CDATA[plot做散点图ggplot2qplot是ggplot2的一个作图函数。]]></description><category></category></item></channel></rss>