<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[duncan]]></title><description><![CDATA[目标不会嫌来的太晚，坚持就一定会成功。]]></description><link>https://blog.youkuaiyun.com/dzjun</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; dzjun]]></copyright><item><title><![CDATA[tensorboard的使用]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/143350973</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/143350973</guid><author>dzjun</author><pubDate>Tue, 29 Oct 2024 21:53:46 +0800</pubDate><description><![CDATA[注：可以tensorboard --logdir='/Users/duncandai/pytorch/logdir' --port=6007设置在6007端口打开。writer = SummaryWriter('logdir') #将生成的文件保存在当前路径的logdir目录下。writer.add_scalar("y=x", i,i) #这里生成一个y=x的函数。4.点击上面的链接，可以打开tensorboard查看。1.安装tensorboard。]]></description><category></category></item><item><title><![CDATA[pytorch环境安装]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/143350773</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/143350773</guid><author>dzjun</author><pubDate>Tue, 29 Oct 2024 21:43:30 +0800</pubDate><description><![CDATA[注：这里默认已经安装anaconda，所以自动带了python, pip3之类，看这里可以看到pytorch需要的python版本为3.9及以上的。4.验证是否安装成功,在命令行import torch，不报错就说明安装成功。2.官网会自动检测当前电脑的环境，生成安装命令。1.请求pytorch官网。3.执行上面的命令进行安装。]]></description><category></category></item><item><title><![CDATA[pycharm配置black进行代码格式化]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/129318515</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/129318515</guid><author>dzjun</author><pubDate>Fri, 03 Mar 2023 13:33:15 +0800</pubDate><description><![CDATA[pycharm代码格式化工具-black的配置]]></description><category></category></item><item><title><![CDATA[git密码忘记]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/129318482</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/129318482</guid><author>dzjun</author><pubDate>Fri, 03 Mar 2023 13:33:08 +0800</pubDate><description><![CDATA[git密码更新]]></description><category></category></item><item><title><![CDATA[history查看历史命令，并显示运行时间]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/129318451</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/129318451</guid><author>dzjun</author><pubDate>Fri, 03 Mar 2023 13:29:56 +0800</pubDate><description><![CDATA[history命令]]></description><category></category></item><item><title><![CDATA[git使用详解]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/129318282</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/129318282</guid><author>dzjun</author><pubDate>Fri, 03 Mar 2023 13:20:13 +0800</pubDate><description><![CDATA[git使用详解：创建分支，克隆项目，合并分支，分支重命名，冲突解决等]]></description><category></category></item><item><title><![CDATA[yum install 无法正常执行]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/129305651</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/129305651</guid><author>dzjun</author><pubDate>Thu, 02 Mar 2023 17:46:47 +0800</pubDate><description><![CDATA[yum install 安装报错error: cannot open Packages database in /var/lib/rpm]]></description><category></category></item><item><title><![CDATA[金融风控指标:WOE, IV, KS，LIFT值和PSI]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/115918569</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/115918569</guid><author>dzjun</author><pubDate>Tue, 20 Apr 2021 21:05:09 +0800</pubDate><description><![CDATA[1.WOE：表示自变量在分组内取值对于目标变量（违约概率）的影响

•WOE的计算公式是：

•WOE=ln[(坏样本/总坏样本)/(好样本/总好样本)]=ln[(坏样本/好样本)/(总坏样本/总好样本)]






Pyi：是这个组中响应客户（坏样本）占所有样本中所有响应客户的比例

Pni：是这个组中未响应客户（好样本）占样本中所有未响应客户的比例

#yi：是这个组中响应客户的数量

#ni：是这个组中未响应客户的数量

#yT：是样本中所有响应客户(总的坏样本)的数量

#nT：是样本中所有未响.]]></description><category></category></item><item><title><![CDATA[jupyter正确使用py文件的姿势]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/88305141</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/88305141</guid><author>dzjun</author><pubDate>Wed, 12 Dec 2018 08:27:07 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[yum: Cannot find a valid baseurl for repo: migsrv解决方法]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/88305140</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/88305140</guid><author>dzjun</author><pubDate>Mon, 03 Sep 2018 10:33:45 +0800</pubDate><description><![CDATA[po=migsrv error was
14: PYCURL ERROR 22 - &quot;The requested URL returned error: 404&quot;
Error: Cannot find a valid baseurl for repo: migsrv


解决办法：删除migsrv.repo文件

rm -rf  /etc/yum.repos.d/migsrv.repo
]]></description><category></category></item><item><title><![CDATA[yum: Cannot find a valid baseurl for repo: migsrv解决方法]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/84925510</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/84925510</guid><author>dzjun</author><pubDate>Mon, 03 Sep 2018 10:33:45 +0800</pubDate><description><![CDATA[yum安装程序报错：

Loaded plugins: fastestmirror
Setting up Update Process
Determining fastest mirrors
Could not retrieve mirrorlist http://yum.wsd.com/index.php?release=1.2&amp;arch=x86_64&amp;repo=migsrv erro...]]></description><category></category></item><item><title><![CDATA[ pandas 计数value_counts()]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/88305139</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/88305139</guid><author>dzjun</author><pubDate>Thu, 23 Aug 2018 10:12:53 +0800</pubDate><description><![CDATA[rue)


参数:	
normalize : boolean, default False　如果设置为true，则以百分比的形式显示
sort : boolean, default True　是否排序
ascending : boolean, default False　默认降序排序
bins : integer, optional 而不是数值计算，把它们分成半开放的箱子，一个方便的pd.cut，只适用于数字数据
dropna : boolean, default True　默认删除na值]]></description><category></category></item><item><title><![CDATA[ pandas 计数value_counts()]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/84925056</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/84925056</guid><author>dzjun</author><pubDate>Thu, 23 Aug 2018 10:12:53 +0800</pubDate><description><![CDATA[在pandas里面常用用value_counts统计某个字段各种值数据出现的频率

pandas.Series.value_counts
Series.value_counts(normalize=False, sort=True, ascending=False, bins=None, dropna=True)


参数:	
normalize : boolean, default...]]></description><category></category></item><item><title><![CDATA[Pandas相关函数用法]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/88305138</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/88305138</guid><author>dzjun</author><pubDate>Tue, 07 Aug 2018 17:32:42 +0800</pubDate><description><![CDATA[None, levels=None, names=None, verify_integrity=False)
1
2
参数说明 
objs: series，dataframe或者是panel构成的序列lsit 
axis： 需要合并链接的轴，0是行，1是列 
join：连接的方式 inner，或者outer,如果join_axes没指定则是使用index来join
keys：用于标识数据来自哪个dataframe

其他一些参数不常用，用的时候再补上说明。


result = pd]]></description><category></category></item><item><title><![CDATA[特征工程]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/88305137</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/88305137</guid><author>dzjun</author><pubDate>Wed, 30 May 2018 11:03:11 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[特征工程]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/84921343</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/84921343</guid><author>dzjun</author><pubDate>Wed, 30 May 2018 11:03:11 +0800</pubDate><description><![CDATA[1.什么是特征工程
　　顾名思义，其本质是一项工程活动，目的是最大限度地从原始数据中提取特征以供算法和模型使用。特征工程是指从原始数据转换为特征向量的过程。

2.特征工程作用
　　在业界广泛流传一句话：数据和特征决定了机器学习的上限，而模型和算法只是逼近这个上限而已。

（1）特征越好，灵活性越强

只要特征选得好，即使是一般的模型（或算法）也能获得很好的性能，因为大多数模型（或...]]></description><category></category></item><item><title><![CDATA[损失函数总结]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/88305136</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/88305136</guid><author>dzjun</author><pubDate>Sun, 20 May 2018 15:18:03 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[损失函数总结]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/84920829</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/84920829</guid><author>dzjun</author><pubDate>Sun, 20 May 2018 15:18:03 +0800</pubDate><description><![CDATA[1.什么是损失函数
损失函数（loss function）是用来估量模型的预测值f(x)与真实值Y的不一致程度，它是一个非负实值函数,通常使用L(Y, f(x))来表示，损失函数越小，模型认为性能就越好。


2.损失函数，代价函数和目标函数的区别
损失函数：计算的是一个样本的误差

代价函数：是整个训练集上所有样本误差的平均

目标函数：代价函数 + 正则化项

目标函数公...]]></description><category></category></item><item><title><![CDATA[激活函数汇总及各优缺点]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/88305135</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/88305135</guid><author>dzjun</author><pubDate>Mon, 14 May 2018 11:05:24 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[激活函数汇总及各优缺点]]></title><link>https://blog.youkuaiyun.com/dzjun/article/details/84920519</link><guid>https://blog.youkuaiyun.com/dzjun/article/details/84920519</guid><author>dzjun</author><pubDate>Mon, 14 May 2018 11:05:24 +0800</pubDate><description><![CDATA[激活函数

1. 什么是激活函数

如下图，在神经元中，输入的 inputs 通过加权，求和后，还被作用了一个函数，这个函数就是激活函数 Activation Function。


[img]http://dl2.iteye.com/upload/attachment/0129/8036/e5a56dfd-c778-3397-9e24-a786b7d9d7a9.png[/img]...]]></description><category></category></item></channel></rss>