<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[what_how_why2020的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/what_how_why2020</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; what_how_why2020]]></copyright><item><title><![CDATA[模型进阶与神经网络]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/151903611</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/151903611</guid><author>what_how_why2020</author><pubDate>Sat, 20 Sep 2025 20:34:31 +0800</pubDate><description><![CDATA[从输入特征数以及激活函数两方面来优化模型，并引出神经网络和深度学习相关概念]]></description><category></category></item><item><title><![CDATA[机器学习基本概述]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/151903547</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/151903547</guid><author>what_how_why2020</author><pubDate>Sat, 20 Sep 2025 20:32:21 +0800</pubDate><description><![CDATA[以观看量预测为例介绍机器学习的基础概念]]></description><category></category></item><item><title><![CDATA[注意力代码实现]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/145839396</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/145839396</guid><author>what_how_why2020</author><pubDate>Mon, 24 Feb 2025 22:24:49 +0800</pubDate><description><![CDATA[【代码】注意力代码实现。]]></description><category></category></item><item><title><![CDATA[论文阅读 Popularity prediction for marketer-generated content A text-guided attention neural network ...]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/145837444</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/145837444</guid><author>what_how_why2020</author><pubDate>Mon, 24 Feb 2025 20:28:54 +0800</pubDate><description><![CDATA[定义若干符号表示，描述问题为给定一组 MGC，提取特征后，学习一个函数去预测流行度得分。假设存在一个包含DDD个营销者生成内容（MGC）的集合。对于索引为iii的 MGC，其由文本描述diwi1wi2⋯win⋯wiNidi​wi1​wi2​⋯win​⋯wiNi​​表示，对应图像集IiI_i=Ii​pi1pi2⋯pim⋯piMipi1​pi2​⋯pi。]]></description><category></category></item><item><title><![CDATA[分词 tokenization]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/145787059</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/145787059</guid><author>what_how_why2020</author><pubDate>Fri, 21 Feb 2025 22:32:32 +0800</pubDate><description><![CDATA[子词（subword）的粒度介于字和词之间，常见的 subword 分词方法有：Byte Pair Encoding(BPE)、Wordpiece 等。这些方法通过统计文本中的字串频率，自动生成一种分词词典，能够有效应对未登录词（OOV）问题，同时保持一定的语义完整性。如果有些文章只提到中国、科学等次，那么当粒度增加，就会有更多的文章被检索到，因为总的数量是固定的，所以召回率就会增加。召回率的计算公式：真正相关的结果（正确找到的相关结果）/真正相关的结果+假负相关的结果（为找到的相关结果）]]></description><category></category></item><item><title><![CDATA[Transformer 模型]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/145716342</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/145716342</guid><author>what_how_why2020</author><pubDate>Tue, 18 Feb 2025 22:26:22 +0800</pubDate><description><![CDATA[本文内容基于 Attention is all you need 论文。]]></description><category></category></item><item><title><![CDATA[文本表示方法]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/145657948</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/145657948</guid><author>what_how_why2020</author><pubDate>Sat, 15 Feb 2025 22:19:15 +0800</pubDate><description><![CDATA[其中，CBOW 是根据上下文来预测中心词，而 Skip-Gram 是根据中心词来预测上下文。如果上下文窗口为 1，对于 CBOW 来说，you say goodbye 中的目标预测词为 say，上下文为 you goodbye。单词在特定文本中的重要性得分表示为：单词在文本出现的频率和出现改单词的文本数量在语料库中的频率。独热编码会根据语料库中的单词个数，来确定词向量的维度。分布式表征，预先确定词向量的维度，生成的词向量。中每个单词在文本中出现的次数来表示文本。，将文本视为词的集合，通过。]]></description><category></category></item><item><title><![CDATA[论文阅读 Lexicon information in neural sentiment analysis: a multi-task learning approach]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/145605380</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/145605380</guid><author>what_how_why2020</author><pubDate>Thu, 13 Feb 2025 09:54:11 +0800</pubDate><description><![CDATA[阅读这篇论文的目的是了解这篇论文如何结合情感词典实现句子级别的情感预测，所以对论文中的「引言」、「文献综述」部分略读。]]></description><category></category></item><item><title><![CDATA[transformer 基础知识]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/145599765</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/145599765</guid><author>what_how_why2020</author><pubDate>Wed, 12 Feb 2025 20:53:23 +0800</pubDate><description><![CDATA[概要：简要记录 Encoder-Decoder 架构、seq2seq 模型、Attention 机制。]]></description><category></category></item><item><title><![CDATA[Neo4j Desktop界面认识以及数据库备份与还原]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/139592441</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/139592441</guid><author>what_how_why2020</author><pubDate>Tue, 11 Jun 2024 10:32:41 +0800</pubDate><description><![CDATA[Neo4j Desktop界面认识以及数据库备份与还原]]></description><category></category></item><item><title><![CDATA[论文阅读 A Distributional Framework for Data Valuation]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/139496302</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/139496302</guid><author>what_how_why2020</author><pubDate>Thu, 06 Jun 2024 11:50:05 +0800</pubDate><description><![CDATA[本论文解决的问题量化数据价值（机器学习模型训练中各个数据点的贡献）避免数据价值受到其所处数据集的影响，使数据点的估值更加稳定、一致。]]></description><category></category></item><item><title><![CDATA[论文阅读 Explainable Image Similarity Integrating Siamese Networks and Grad-CAM]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/139410737</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/139410737</guid><author>what_how_why2020</author><pubDate>Mon, 03 Jun 2024 11:50:27 +0800</pubDate><description><![CDATA[给出论文（Explainable Image Similarity Integrating Siamese Networks and Grad-CAM）的内容解读、代码运行说明。]]></description><category></category></item><item><title><![CDATA[win11 Terminal 部分窗口美化]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/138325978</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/138325978</guid><author>what_how_why2020</author><pubDate>Mon, 29 Apr 2024 21:24:34 +0800</pubDate><description><![CDATA[美化 win11 Terminal 部分终端窗口：cmd、anaconda prompt
]]></description><category></category></item><item><title><![CDATA[Data Shapley Value 笔记]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/137454369</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/137454369</guid><author>what_how_why2020</author><pubDate>Sun, 07 Apr 2024 10:38:34 +0800</pubDate><description><![CDATA[本文为 Data Shapley: Equitable Valuation of Data for Machine Learning 的阅读笔记，涉及论文中的 Data Shapley Value 计算公式、两种实现算法、实验应用部分的梳理。

为理解 Data Shapley Value，本文首先讨论 Shapley Value的相关内容，利用一个具体实例计算 Shapley Value 并讨论其计算公式。而后，解释一脉相承的 Data Shapley Value 计算公式、两种实现算法的伪代码。]]></description><category></category></item><item><title><![CDATA[标准的README]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/135825851</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/135825851</guid><author>what_how_why2020</author><pubDate>Wed, 24 Jan 2024 16:37:10 +0800</pubDate><description><![CDATA[GitHub 中标准的 README.md 写法]]></description><category></category></item><item><title><![CDATA[GitFlow工作流]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/135736939</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/135736939</guid><author>what_how_why2020</author><pubDate>Sun, 21 Jan 2024 22:43:59 +0800</pubDate><description><![CDATA[基于 Git 这一版本控制系统，通过定义不同的分支，探索合适的工作流程来完成开发、测试、修改等方面的需求。]]></description><category></category></item><item><title><![CDATA[Git 标签管理]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/135684362</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/135684362</guid><author>what_how_why2020</author><pubDate>Thu, 18 Jan 2024 21:14:38 +0800</pubDate><description><![CDATA[简述 Git 标签管理中的常用命令]]></description><category></category></item><item><title><![CDATA[Git 分支管理]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/135684306</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/135684306</guid><author>what_how_why2020</author><pubDate>Thu, 18 Jan 2024 21:13:42 +0800</pubDate><description><![CDATA[简述 Git 分支管理中的常用命令以及处理冲突分支、Bug 分支、多人协作中分支管理等问题。]]></description><category></category></item><item><title><![CDATA[Git 远程仓库]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/135684257</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/135684257</guid><author>what_how_why2020</author><pubDate>Thu, 18 Jan 2024 21:10:24 +0800</pubDate><description><![CDATA[以 GitHub/Gitee 为例，作为简述 Git 中的远程仓库，以及常用命令。]]></description><category></category></item><item><title><![CDATA[Git 基本命令与操作流]]></title><link>https://blog.youkuaiyun.com/what_how_why2020/article/details/135684207</link><guid>https://blog.youkuaiyun.com/what_how_why2020/article/details/135684207</guid><author>what_how_why2020</author><pubDate>Thu, 18 Jan 2024 21:08:03 +0800</pubDate><description><![CDATA[记录 Git 中的基本命令和创建仓库、提交文件、删除文件等方面的操作]]></description><category></category></item></channel></rss>