<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[SpicyCoder的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/SpicyCoder</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; SpicyCoder]]></copyright><item><title><![CDATA[Linux服务器设置ssh私钥登录]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/122733113</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/122733113</guid><author>SpicyCoder</author><pubDate>Fri, 28 Jan 2022 19:40:52 +0800</pubDate><description><![CDATA[Linux服务器创建用户与私钥

服务器居然被黑客通过暴力ssh破解并进行挖矿，本篇文章将讲述如何创建用户私钥，并关闭ssh密码登录

创建新用户并创建私钥
使用root用户创建新用户，名为xxx，并禁用密码登录。
指定用户使用的shell解释器为bash，指定用户主目录路径为/home/xxx或者其他自定义路径。
adduser xxx --disabled-password -s /bin/bash -d /home/xxx

对已存在的用户创建私钥
首先使用root权限，切换到用户xxx，进入该用户x]]></description><category></category></item><item><title><![CDATA[[论文阅读]Deep Relational Reasoning Graph Network for Arbitrary Shape Text Detection]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/105072570</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/105072570</guid><author>SpicyCoder</author><pubDate>Fri, 27 Mar 2020 11:50:40 +0800</pubDate><description><![CDATA[[论文阅读]Deep Relational Reasoning Graph Network for Arbitrary Shape Text Detection
用于任意形状文本检测的深度关系推理图网络
文章收录于2020 CVPR
[论文地址]https://arxiv.org/abs/2003.07493
[代码地址]https://github.com/GXYM/DRRG
文章目录摘要简介方...]]></description><category></category></item><item><title><![CDATA[[论文阅读]Chinese Street View Text: Large-scale Chinese Text Reading with Partially Supervised Learning]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/104928424</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/104928424</guid><author>SpicyCoder</author><pubDate>Fri, 20 Mar 2020 11:01:47 +0800</pubDate><description><![CDATA[文章目录摘要简介相关工作文本阅读基准端到端文本阅读弱监督和部分监督学习方法实验参考文献
文章被收录于ICCV2019
[论文地址]http://openaccess.thecvf.com/content_ICCV_2019/html/Sun_Chinese_Street_View_Text_Large-Scale_Chinese_Text_Reading_With_Partially_ICCV_2...]]></description><category></category></item><item><title><![CDATA[[论文阅读]TextDragon: An End-to-End Framework for Arbitrary Shaped Text Spotting阅读笔记]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/104774201</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/104774201</guid><author>SpicyCoder</author><pubDate>Thu, 12 Mar 2020 17:34:45 +0800</pubDate><description><![CDATA[TextDragon: An End-to-End Framework for Arbitrary Shaped Text Spotting阅读笔记
文章被收录于ICCV2019
[论文地址]:http://openaccess.thecvf.com/content_ICCV_2019/html/Feng_TextDragon_An_End-to-End_Framework_for_Arbitra...]]></description><category></category></item><item><title><![CDATA[Caffe代码迁移为PyTorch的注意事项]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/103498480</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/103498480</guid><author>SpicyCoder</author><pubDate>Mon, 16 Dec 2019 20:09:12 +0800</pubDate><description><![CDATA[进入研究生生活已经有三个月了，每天的生活就是上课、看论文、跑模型，也从来没有自己动手写过完整的模型。最近想在一个Caffe模型上加个模块，奈何实在改不动Caffe代码（C++忘的差不多了，日后一定好好看看Caffe底层），于是转战PyTorch。
本文主要介绍一下注意事项，不具体展开介绍怎么样一步一步来，只说明一下大致步骤，具体细节只有做了才能明白。
Caffe网络结构的迁移

参考文章
PyTo...]]></description><category></category></item><item><title><![CDATA[cs231n assignment2 PyTorch]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/99415480</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/99415480</guid><author>SpicyCoder</author><pubDate>Wed, 21 Aug 2019 10:37:41 +0800</pubDate><description><![CDATA[文章目录Barebones PyTorchThree-Layer ConvNetTraining a ConvNetPyTorch Module API
Barebones PyTorch
Three-Layer ConvNet
使用pytorch抽象等级1的方式实现卷积神经网络。
three_layer_convnet()
out1 = F.conv2d(x, conv_w1, bias=con...]]></description><category></category></item><item><title><![CDATA[cs231n assignment2 ConvolutionalNetworks]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/98749292</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/98749292</guid><author>SpicyCoder</author><pubDate>Sun, 11 Aug 2019 20:49:58 +0800</pubDate><description><![CDATA[文章目录ConvolutionNaive forward passNaive backward pass
Convolution
Naive forward pass
实现卷积操作，按照卷积的流程，写出最能理解的代码。
N, C, H, W = x.shape
F, _, HH, WW = w.shape
stride, pad = conv_param['stride'], conv_param...]]></description><category></category></item><item><title><![CDATA[cs231n assignment2 dropout]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/98208764</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/98208764</guid><author>SpicyCoder</author><pubDate>Sat, 03 Aug 2019 11:08:52 +0800</pubDate><description><![CDATA[代码：https://github.com/LiuZhe6/CS231N
为了防止神经网络过拟合数据，可以采用dropout方法。其主要思想是：对隐藏层中部分输出或者权重随机置为0。
文章目录Dropoutforward passbackward passInline Question
Dropout
forward pass
题目要求使用inverted dropout，其主要思想是在训练阶段在...]]></description><category></category></item><item><title><![CDATA[cs231n assignment2 Batch Normalization]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/97796858</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/97796858</guid><author>SpicyCoder</author><pubDate>Fri, 02 Aug 2019 10:30:21 +0800</pubDate><description><![CDATA[BatchNormalization的反向推导比之前的稍微复杂一些，但是画出计算图后，从后往前推导就会变的简单。
文章目录Batch normalizationforwardbackwardLayer Normalization: ImplementationInline Question
Batch normalization
forward
首先实现layers.py中的batchnorm_f...]]></description><category></category></item><item><title><![CDATA[cs231n assignment2 Fully-connected Neural Network]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/96830721</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/96830721</guid><author>SpicyCoder</author><pubDate>Mon, 29 Jul 2019 11:35:36 +0800</pubDate><description><![CDATA[开始了卷积神经网络的学习，这部分内容还是非常多的。本次作业分为五个部分：Q1：Fully-connected Neural Network、Q2：Batch Normalization、Q3：Dropout、Q4:Convolutional Networks、Q5：PyTorch / TensorFlow on CIFAR-10。
以下内容为全连接网络内容，采用模块化的设计，思路非常清晰。
准备工...]]></description><category></category></item><item><title><![CDATA[cs231n assignment1 features]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/96323584</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/96323584</guid><author>SpicyCoder</author><pubDate>Wed, 17 Jul 2019 17:25:28 +0800</pubDate><description><![CDATA[features
前几个作业都是直接将图片的原始像素作为模型输入，本次作业是通过使用定向梯度直方图Histogram of Oriented Gradients (HOG)和HSV颜色空间。简单说，HOG不考虑图片的颜色信息，只捕获图片的纹理；而颜色直方图只考虑颜色，不考虑纹理。使用这两个方法一起工作，效果会更好。
Train SVM on features
learning_rates = [0...]]></description><category></category></item><item><title><![CDATA[cs231n assignment1 two-layer-net]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/95903137</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/95903137</guid><author>SpicyCoder</author><pubDate>Tue, 16 Jul 2019 11:36:09 +0800</pubDate><description><![CDATA[two-layer-net
首先完成神经网络对scores和损失函数的计算，其中激活函数使用RELU函数，即max(0,x)函数。
neural_net.py的loss()函数
# *****START OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****
h1 = np.maximum(0,X.dot(W1) + b1)
scores = h1.d...]]></description><category></category></item><item><title><![CDATA[cs231n assignment1 softmax]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/95588287</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/95588287</guid><author>SpicyCoder</author><pubDate>Fri, 12 Jul 2019 11:48:38 +0800</pubDate><description><![CDATA[Softmax
Softmax的损失函数为
Li=−log⁡pyi=−log⁡(efyi∑jefj)=−fyi+log⁡∑jefjL_{i}=-\log p_{y_{i}}=-\log \left(\frac{e^{f_{y_{i}}}}{\sum_{j} e^{f_{j}}}\right)=-f_{y_{i}}+\log \sum_{j} e^{f_{j}}Li​=−logpyi​​=−log(...]]></description><category></category></item><item><title><![CDATA[cs231n assignment1 SVM]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/95351120</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/95351120</guid><author>SpicyCoder</author><pubDate>Wed, 10 Jul 2019 17:20:29 +0800</pubDate><description><![CDATA[上一篇：cs231n assignment1 knn
文章目录SVMInline Question
SVM
支持向量机的损失函数为
Li=∑j!=yimax⁡(0,sj−syi+△)L_{i}=\sum_{j!=y_{i}} \max \left(0, s_{j}-s_{y_{i}}+\triangle\right)Li​=j!=yi​∑​max(0,sj​−syi​​+△)
其中sjs_{j}s...]]></description><category></category></item><item><title><![CDATA[cs231n assignment1 knn]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/94992552</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/94992552</guid><author>SpicyCoder</author><pubDate>Tue, 09 Jul 2019 14:37:30 +0800</pubDate><description><![CDATA[文章目录准备阶段Assignment11.  k-Nearest Neighbor classifier2. SVM3. Softmax classifier4. Two-Layer Neural Network5. Higher Level Representations: Image FeaturesInline ProblemInline Question 1Inline Question ...]]></description><category></category></item><item><title><![CDATA[吴恩达机器学习第十一周测验]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/94608735</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/94608735</guid><author>SpicyCoder</author><pubDate>Thu, 04 Jul 2019 13:42:46 +0800</pubDate><description><![CDATA[测验：Application: Photo OCR
第一题
Suppose you are running a sliding window detector to find
text in images. Your input images are 1000x1000 pixels. You
will run your sliding windows detector at two scales...]]></description><category></category></item><item><title><![CDATA[吴恩达机器学习第十周测试]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/94456706</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/94456706</guid><author>SpicyCoder</author><pubDate>Tue, 02 Jul 2019 17:19:57 +0800</pubDate><description><![CDATA[第一题

答案
B
分析：当代价函数呈上升趋势的时候，可以试着将学习率减小
第二题

答案
CD
分析：
A：随机梯度下降并不能并行化，错误。
B：批量梯度下降是在每一次迭代后计算代价函数，错误。
C：在随机梯度下降算法执行之前，先要将样本打乱，正确。
D：在大量数据样本的情况下，随机梯度下降要比批量梯度下降效率高，正确。
第三题

答案
AD
第四题

答案
CD
第五题

答案
ACD

...]]></description><category></category></item><item><title><![CDATA[吴恩达机器学习第九周测试及编程练习]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/93378224</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/93378224</guid><author>SpicyCoder</author><pubDate>Sat, 29 Jun 2019 19:06:13 +0800</pubDate><description><![CDATA[文章目录测验一：Anomaly Detection第一题第二题第三题第四题第五题测验二：编程练习：
测验一：Anomaly Detection
第一题

答案
AD
分析：
异常检测通常是在大部分数据是正常的情况下，发现少量不同寻常的数据。
第二题

答案
B
分析：当算法检测的结果丢掉了异常数据的时候，此时增加sigma可以减少丢失。
第三题

答案
A
第四题

答案
AD
分析：
A：当使用...]]></description><category></category></item><item><title><![CDATA[吴恩达机器学习第八周测试及编程练习]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/91405029</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/91405029</guid><author>SpicyCoder</author><pubDate>Sat, 15 Jun 2019 10:44:42 +0800</pubDate><description><![CDATA[文章目录测试：Principal Component Analysis第一题第二题第三题第四题第五题编程练习：K-Means Clustering and PCA作业一：Find Closest Centroids (k-Means)作业二：Compute Centroid Means (k-Means)作业三：PCA作业四：Project Data (PCA)作业五：Recover Data (...]]></description><category></category></item><item><title><![CDATA[吴恩达机器学习第七周测试和编程作业]]></title><link>https://blog.youkuaiyun.com/SpicyCoder/article/details/90598527</link><guid>https://blog.youkuaiyun.com/SpicyCoder/article/details/90598527</guid><author>SpicyCoder</author><pubDate>Tue, 28 May 2019 20:21:18 +0800</pubDate><description><![CDATA[文章目录测验：Support Vector Machines第一题第二题第三题第四题第五题编程作业：Support Vector Machines作业一：Gaussian Kernel作业二：Parameters (C, sigma) for Dataset 3作业三：Email Preprocessing作业四：Email Feature Extraction
测验：Support Vector...]]></description><category></category></item></channel></rss>