<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[python0930233的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/python0930233</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; python0930233]]></copyright><item><title><![CDATA[深度学习扩展]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/145903662</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/145903662</guid><author>python0930233</author><pubDate>Thu, 27 Feb 2025 19:50:14 +0800</pubDate><description><![CDATA[—把图片离散化为一排向量将图片拆为若干patch，每个patch可以看作一个字，再加上位置信息就可以输入到transformer中。ViT的核心是将图像视为一系列局部块的序列（类似自然语言中的单词），通过Transformer的全局自注意力机制捕捉图像中的长距离依赖关系。将输入图像（如224×224）分割为固定大小的非重叠块（例如16×16像素），每个块展平为向量。例如，224×224的图像会生成196个（14×14）块。]]></description><category></category></item><item><title><![CDATA[Bert实战—听课笔记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/145882149</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/145882149</guid><author>python0930233</author><pubDate>Thu, 27 Feb 2025 11:46:07 +0800</pubDate><description><![CDATA[这是下游的文字分类任务。bert就是一个特征提取器，在上游进行MLM和NSP预训练，之后bert就可以从句子中提取特征，做分类任务。]]></description><category></category></item><item><title><![CDATA[1.16学习日记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/145183872</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/145183872</guid><author>python0930233</author><pubDate>Thu, 16 Jan 2025 22:12:09 +0800</pubDate><description><![CDATA[今天学习的是第七节，自然语言处理与bert。]]></description><category></category></item><item><title><![CDATA[1.14学习日记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/145139022</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/145139022</guid><author>python0930233</author><pubDate>Tue, 14 Jan 2025 15:48:37 +0800</pubDate><description><![CDATA[今天学的是第六节：特征提取和无监督学习。不能再玩了……]]></description><category></category></item><item><title><![CDATA[1.12学习日记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/145092402</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/145092402</guid><author>python0930233</author><pubDate>Sun, 12 Jan 2025 20:21:18 +0800</pubDate><description><![CDATA[前两天学校有课加上要回家所以没学习，今天开始继续学习(●ˇ∀ˇ●)今天学第五节下：图片分类半监督。]]></description><category></category></item><item><title><![CDATA[1.9学习日记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/145033238</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/145033238</guid><author>python0930233</author><pubDate>Thu, 09 Jan 2025 15:55:22 +0800</pubDate><description><![CDATA[今天继续学习食物分类图片分类的模型最重要的就是设计维度变化。]]></description><category></category></item><item><title><![CDATA[1.8学习日记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/145018501</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/145018501</guid><author>python0930233</author><pubDate>Thu, 09 Jan 2025 00:02:13 +0800</pubDate><description><![CDATA[今天学习的是第五节上 图像分类实战——食物分类目的：通过模型预测图片是哪类食物。]]></description><category></category></item><item><title><![CDATA[EXCEL基础操作]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/144951869</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/144951869</guid><author>python0930233</author><pubDate>Wed, 08 Jan 2025 18:52:04 +0800</pubDate><description><![CDATA[1.养成好习惯，对源数据要备份2.看量级3.ctrl+shift+L：快速筛选4.1.：UV指的是在一定时间范围内访问某个网站或页面的独立用户数。。用途：UV反映了网站的独立访问人数，因此它能帮助网站分析其用户群体的大小。这个指标是衡量网站吸引力、用户活跃度和潜在市场规模的重要指标。计算方法：通常通过用户的IP地址、浏览器标识符或者登录账户等信息来识别独立用户。4.2.：PV指的是网站页面被浏览的次数。。]]></description><category></category></item><item><title><![CDATA[1.6学习日记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/144970130</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/144970130</guid><author>python0930233</author><pubDate>Mon, 06 Jan 2025 21:10:47 +0800</pubDate><description><![CDATA[1.Conv2d就是卷积，其参数依次为：输入特征图的数量、输出特征图数量（即卷积核数量）、卷积核大小、步长、padding。4.归一化：避免受数据量纲影响，让模型关注数据的分布，保持学习有效性，缓解梯度消失和梯度爆炸。1.rule：计算方便、可以在一定程度上缓解梯度消失。2.pool（3，2）表示3格一pool，每次走2格。一、一般将softmax()的输出视作模型预测的概率分布。需要的参数：预测的分布、真实值的下标。2.用小卷积核代替大的卷积核。二、分类任务的Loss：交叉熵损失。]]></description><category></category></item><item><title><![CDATA[认识数据分析]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/144951671</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/144951671</guid><author>python0930233</author><pubDate>Sun, 05 Jan 2025 22:00:25 +0800</pubDate><description><![CDATA[2、找到数据之间的因果关系。二）要明确业务目标→拆解业务目标。数据分析：观测、实验、应用。]]></description><category></category></item><item><title><![CDATA[1.5学习日记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/144950460</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/144950460</guid><author>python0930233</author><pubDate>Sun, 05 Jan 2025 21:53:14 +0800</pubDate><description><![CDATA[I：输入特征图大小；P：padding；S：步长stride。2.怎么让特征图变小？【不常用，会丢失信息且引入计算】一、今天学了分类任务。]]></description><category></category></item><item><title><![CDATA[1.4学习日记]]></title><link>https://blog.youkuaiyun.com/python0930233/article/details/144935565</link><guid>https://blog.youkuaiyun.com/python0930233/article/details/144935565</guid><author>python0930233</author><pubDate>Sat, 04 Jan 2025 21:42:14 +0800</pubDate><description><![CDATA[能够预测到好的y。根据预测值y‘与真实值y的差值不断训练模型，使最终能够预测得到比较好的y。一、把实战班第三节下终于看完了，但是代码也是似懂非懂的。]]></description><category></category></item></channel></rss>