<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[鼹鼠的胡须 的博客]]></title><description><![CDATA[所有目的地]]></description><link>https://blog.youkuaiyun.com/qq_23968185</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; qq_23968185]]></copyright><item><title><![CDATA[Pytorch中网络不同参数组使用不同优化器]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/122129422</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/122129422</guid><author>qq_23968185</author><pubDate>Fri, 24 Dec 2021 15:40:38 +0800</pubDate><description><![CDATA[Pytorch 优化器，不同参数组使用不同优化器进行参数更新]]></description><category></category></item><item><title><![CDATA[Pytorch定义网络时循环创建变量名]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/119462876</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/119462876</guid><author>qq_23968185</author><pubDate>Fri, 24 Dec 2021 15:20:57 +0800</pubDate><description><![CDATA[Pytorch定义网络时循环创建变量名及pytorch的容器]]></description><category></category></item><item><title><![CDATA[Shell脚本出现$‘\r‘: command not found]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/119176844</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/119176844</guid><author>qq_23968185</author><pubDate>Wed, 28 Jul 2021 12:38:38 +0800</pubDate><description><![CDATA[Shell脚本出现$'\r': command not found问题描述：解决方案
问题描述：
在Windows下编写好sh文件后，在Linux下运行会报错：bash: $’\r’: command not found
解决方案
使用vim打开文件，然后使用命令:set ff=unix，保存文件

...]]></description><category></category></item><item><title><![CDATA[Pyroch中nn.Sequential的多输入问题]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/108277724</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/108277724</guid><author>qq_23968185</author><pubDate>Fri, 28 Aug 2020 13:20:48 +0800</pubDate><description><![CDATA[Pyroch中nn.Sequential的多输入问题nn.Sequential为何不能处理多输入？nn.Sequential 通过单参数处理多输入问题
nn.Sequential为何不能处理多输入？
先上两个小例子直观感受一下：
class MyLayer(nn.Module):
    def __init__(self):
        super(MyLayer, self).__init__()
        self.conv1 = nn.Conv2d(3, 32, 1, 1, bias=Fa]]></description><category></category></item><item><title><![CDATA[win10远程控制ubuntu16.04]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/83538394</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/83538394</guid><author>qq_23968185</author><pubDate>Tue, 30 Oct 2018 10:54:11 +0800</pubDate><description><![CDATA[使用win10远程控制ubuntu16.041、打开终端，安装xrdp,vncserver2、安装desktop sharing3、安装dconf-editor，取消权限限制
使用win10远程控制ubuntu16.04，网上很多需要安装xfce桌面的。今天介绍一下，不需要安装其他桌面，使用Ubuntu16.04自带桌面，漂亮美观。
Ubuntu16.04端：
1、打开终端，安装xrdp,vncs...]]></description><category></category></item><item><title><![CDATA[二分类模型评估之 ROC曲线和PR曲线]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/77826654</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/77826654</guid><author>qq_23968185</author><pubDate>Sun, 03 Sep 2017 22:11:49 +0800</pubDate><description><![CDATA[ROC曲线：ROC曲线一系列根据不同的二分类方式（阈值或分界值），以真阳性率TPR（灵敏度）为纵坐标，假阳性率FPR（1-特异度）为横坐标绘制的曲线。]]></description><category></category></item><item><title><![CDATA[Python数据存储之 h5py]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/77671726</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/77671726</guid><author>qq_23968185</author><pubDate>Tue, 29 Aug 2017 10:27:21 +0800</pubDate><description><![CDATA[1、Python数据存储（压缩）（1）numpy.save ,  numpy.savez , scipy.io.savemat 
numpy和scipy内建的数据存储方式。 
（2）cPickle + gzip 
cPickle是pickle内建的数据存储方式，gzip是常用的文件压缩模块。 
（3）h5py 
h5py是对HDF5文件格式进行读写的python包，关于h5py更多介绍与安装，参考官]]></description><category></category></item><item><title><![CDATA[VS中常见报错及解决方法]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/75096238</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/75096238</guid><author>qq_23968185</author><pubDate>Fri, 14 Jul 2017 09:43:29 +0800</pubDate><description><![CDATA[报错： Error 1 error C2440: ‘initializing’ : cannot convert from ‘wchar_t ’ to ‘char ‘
解决方法：右击工程名 —〉Property—〉General中的character set中选择not set]]></description><category></category></item><item><title><![CDATA[DL去燥之 DnCNNs：Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/72944043</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/72944043</guid><author>qq_23968185</author><pubDate>Fri, 09 Jun 2017 00:02:10 +0800</pubDate><description><![CDATA[网络结构：DnCNNs： feed-forward denoising convolutional neural networks 
学习策略Residual Learning：通过残差学习策略，在网络的隐层隐式地移除干净图片（clean image）。即输入为带噪的观测图片（noisy image），输出为移除了干净图片的残差（噪声）图片（residual image）。这样做的动机是，用残差学习]]></description><category></category></item><item><title><![CDATA[EM算法MATLAB代码及详细注解]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/70940197</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/70940197</guid><author>qq_23968185</author><pubDate>Fri, 28 Apr 2017 19:57:18 +0800</pubDate><description><![CDATA[下面代码为PRML所附的基于混合高斯（MoG）的代码，个人认为编码可读性和风格都值得借鉴。function [label, model, llh] = mixGaussEm(X, init)
% Perform EM algorithm for fitting the Gaussian mixture model.
% Input: 
%   X: d x n data matrix
%   ini]]></description><category></category></item><item><title><![CDATA[矩阵分解（MF）方法及代码]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/70477613</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/70477613</guid><author>qq_23968185</author><pubDate>Sat, 22 Apr 2017 21:57:24 +0800</pubDate><description><![CDATA[一、基于投影梯度法的非负矩阵分解论文：Projected gradient methods for non-negative matrix factorization 
代码：Matlab及Python源码二、基于类牛顿法的最小二乘矩阵近似解法论文：Fast Newton-type Methods for the Least Squares Nonnegative Matrix 
Approxima]]></description><category></category></item><item><title><![CDATA[视频前背景分离论文之（5） Robust PCA via Nonconvex Rank Approximation]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/53690603</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/53690603</guid><author>qq_23968185</author><pubDate>Fri, 16 Dec 2016 11:49:20 +0800</pubDate><description><![CDATA[1、AbstractRPCA是低秩矩阵恢复的general的模型框架， 在某些假设下它能够以很大的概率很好地恢复原矩阵潜在的低秩矩阵（the underlying true low rank matrix）的，然而，实际应用中这些假设并不一定成立 。在RPCA中核范数是通过将奇异值加和来逼近矩阵的秩的，即相当于求奇异值的ℓ1−norm\ell_{1}-norm. 这样得到的近似误差是不可忽略的，因此]]></description><category></category></item><item><title><![CDATA[视频前背景分离论文之（4） A Probabilistic Approach to Robust Matrix Factorization]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/53415033</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/53415033</guid><author>qq_23968185</author><pubDate>Wed, 30 Nov 2016 22:48:09 +0800</pubDate><description><![CDATA[1、AbstractPRMF：formulated with a Laplace error and a Gaussian prior which correspond to an l1loss and an l2regularizer.]]></description><category></category></item><item><title><![CDATA[视频前背景分离论文之（3） Group Sparsity in Nonnegative Matrix Factorization]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/53414872</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/53414872</guid><author>qq_23968185</author><pubDate>Wed, 30 Nov 2016 22:32:48 +0800</pubDate><description><![CDATA[1、Abstract（1）features or data items belonging to the same group share the same sparsity pattern in low-rank factors. 
（2）Proposed \textcolor{blue}{mixed-norm} regularization to promote group sparsity i]]></description><category></category></item><item><title><![CDATA[视频前背景分离论文之（2） GOSUS: Grassmannian Online Subspace Updates with Structured-sparsity]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/53414748</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/53414748</guid><author>qq_23968185</author><pubDate>Wed, 30 Nov 2016 22:23:08 +0800</pubDate><description><![CDATA[1、Abstract:(1)  view online estimation procedure as an approximate optimization on a Grassmannian(格拉斯曼流形); 
(2) leverage structured-sparsity to model both homogeneous perturbations of the subspace and]]></description><category></category></item><item><title><![CDATA[视频前背景分离论文之（1） Online Robust PCA via Stochastic Optimization]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/53414508</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/53414508</guid><author>qq_23968185</author><pubDate>Wed, 30 Nov 2016 22:05:27 +0800</pubDate><description><![CDATA[视频前后背景分离论文：Online Robust PCA via Stochastic Optimization]]></description><category></category></item><item><title><![CDATA[ISTA算法求解L1正则化问题]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/53414225</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/53414225</guid><author>qq_23968185</author><pubDate>Wed, 30 Nov 2016 21:41:11 +0800</pubDate><description><![CDATA[ISTA算法求解L1正则化问题]]></description><category></category></item><item><title><![CDATA[SLIC原理简介及VLFeat代码示例]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/52878331</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/52878331</guid><author>qq_23968185</author><pubDate>Thu, 20 Oct 2016 23:25:27 +0800</pubDate><description><![CDATA[SLIC，即简单线性迭代聚类，是一种简单有效的图像分割方法，其本质是局部空间的K-means聚类。]]></description><category></category></item><item><title><![CDATA[机器学习之 特征选择]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/52510992</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/52510992</guid><author>qq_23968185</author><pubDate>Mon, 12 Sep 2016 19:13:27 +0800</pubDate><description><![CDATA[机器学习过程中，特征选择是避免维数灾难和减低学习难度的有效途径。特征选择的过程，即为选择对当前学习任务有用的属性（相关特征）和去除无用属性（冗余特征）的过程。那么，如何选择有用的特征子集以及如何评价相对好的特征子集是特征选择的关键步骤。]]></description><category></category></item><item><title><![CDATA[感知机算法原理及代码实现]]></title><link>https://blog.youkuaiyun.com/qq_23968185/article/details/52494943</link><guid>https://blog.youkuaiyun.com/qq_23968185/article/details/52494943</guid><author>qq_23968185</author><pubDate>Sat, 10 Sep 2016 14:44:48 +0800</pubDate><description><![CDATA[最原始的感知机学习算法仅仅对于线性可分的训练集是收敛的.]]></description><category></category></item></channel></rss>