<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[qq_43456497的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/qq_43456497</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; qq_43456497]]></copyright><item><title><![CDATA[vscode配置Latex+SumatraPDF]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/136812242</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/136812242</guid><author>qq_43456497</author><pubDate>Mon, 18 Mar 2024 16:45:59 +0800</pubDate><description><![CDATA[原本使用Overleaf写毕业论文，最近发现内容过多时会出错，项目超过了我们免费计划的编译超时限制，因此考虑使用vscode配置latex环境，为了显示方便，使用SumatraPDF查看器。]]></description><category></category></item><item><title><![CDATA[大模型微调总结1-总览]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/135072997</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/135072997</guid><author>qq_43456497</author><pubDate>Wed, 27 Dec 2023 11:12:04 +0800</pubDate><description><![CDATA[参数高效微调（Parameter-efficient fine-tuning）是指微调少量或额外的模型参数，固定大部分预训练模型（LLM）参数，从而大大降低了计算和存储成本，同时，也能实现与全量参数微调相当的性能。参数高效微调方法甚至在某些情况下比全量微调效果更好，可以更好地泛化到域外场景。这也是现在主流的微调方法。高效微调技术可以粗略分为以下三大类：增加额外参数（A）、选取一部分参数更新（S）、引入重参数化（R）。]]></description><category></category></item><item><title><![CDATA[详解知识蒸馏原理和代码]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/130853581</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/130853581</guid><author>qq_43456497</author><pubDate>Wed, 24 May 2023 20:32:06 +0800</pubDate><description><![CDATA[知识蒸馏是一种模型压缩的有效解决方案；总的来说，学生模型通过蒸馏训练来获取教师知识，小模型学习到了大模型的泛化能力，保留了大模型的性能，同时降低了模型的大小和复杂性，模型更轻量易于部署。]]></description><category></category></item><item><title><![CDATA[Pytorch中易混淆的损失函数与激活函数【softmax, log_softmax, NLLLoss, CrossEntropy】]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/130853961</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/130853961</guid><author>qq_43456497</author><pubDate>Wed, 24 May 2023 19:53:20 +0800</pubDate><description><![CDATA[softmax：将一个数值序列映射到概率空间（每个元素分布并且所有和为1）：在softmax的基础上取对数NLLLoss：对log_softmax与one-hot进行计算：衡量两个概率分布的差别（交叉熵）在分类问题中，CrossEntropy等价于log_softmax 结合 nll_loss。]]></description><category></category></item><item><title><![CDATA[深入理解2D卷积和3D卷积]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/130713857</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/130713857</guid><author>qq_43456497</author><pubDate>Tue, 16 May 2023 21:03:09 +0800</pubDate><description><![CDATA[在项目中用到了conv3但是对其背后的原理还有一些模糊的地方，conv2d与多通道的conv2d的区别在哪里？conv3d的思想理论是什么？对此进行探究和记录...... 
首先要明确多通道的2d卷积和3d卷积是不一样的，3d是可以在通道中移动的，2d不可以]]></description><category></category></item><item><title><![CDATA[I3D--视频理解必读论文总结]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/130706379</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/130706379</guid><author>qq_43456497</author><pubDate>Tue, 16 May 2023 20:46:02 +0800</pubDate><description><![CDATA[在视频理解领域，I3D是一篇不得不读的文章，在他之前，当时人的解决方案大多数是以双流网络为主进行视频分类。在UCF-101和HMDB-51数据集上进行测试对比结果，当I3D提出之后，基本上将UCF-101和HMDB-51数据集刷到顶峰了，同时，使得3DCNN在视频理解领域大放异彩，直到近几年VIT的出现，才有新的方向对视频理解领域进行推进；I3D也提出了一个新的数据集--Kinetics Dataset，之后的方法都需要在这个数据集上进行对比。]]></description><category></category></item><item><title><![CDATA[机器学习中的相关概念【监督，半监督，无监督，弱监督，自监督】]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/130692628</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/130692628</guid><author>qq_43456497</author><pubDate>Mon, 15 May 2023 20:57:30 +0800</pubDate><description><![CDATA[机器学习中大体可以把任务分为监督学习，无监督学习。在我们的学习和资料查阅中会出现，半监督，自监督，弱监督这样的词汇，让人心生疑惑，在此进行一个梳理。以上各个概念的分类并不是严格互斥的。]]></description><category></category></item><item><title><![CDATA[深入了解梯度消失与梯度爆炸]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/130604158</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/130604158</guid><author>qq_43456497</author><pubDate>Fri, 12 May 2023 20:36:11 +0800</pubDate><description><![CDATA[本文探讨深度学习中经常会提到的概念–梯度消失与梯度爆炸。他们是影响模型收敛，学习好坏的一个重要因素，对此现象也提出了对应的解决方案。在此记录其概念，原因和相关的解决方案，仅供参考。]]></description><category></category></item><item><title><![CDATA[总结常见评价指标]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/130580126</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/130580126</guid><author>qq_43456497</author><pubDate>Tue, 09 May 2023 20:50:13 +0800</pubDate><description><![CDATA[整理一下在机器学习中常见的评价指标，包括：
- 混淆矩阵，TPR，FPR，TNR，FNR；
- Precision，Recall，Accuracy，F-score(F1-meature)
- ROC曲线，AUC；]]></description><category></category></item><item><title><![CDATA[Leetcode 238. 除自身以外数组的乘积]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129352384</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129352384</guid><author>qq_43456497</author><pubDate>Sun, 05 Mar 2023 22:41:59 +0800</pubDate><description><![CDATA[中等题。]]></description><category></category></item><item><title><![CDATA[Leetcode 494. 目标和]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129350470</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129350470</guid><author>qq_43456497</author><pubDate>Sun, 05 Mar 2023 22:21:32 +0800</pubDate><description><![CDATA[中等题。]]></description><category></category></item><item><title><![CDATA[Leetcode 279. 完全平方数]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129350168</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129350168</guid><author>qq_43456497</author><pubDate>Sun, 05 Mar 2023 20:20:50 +0800</pubDate><description><![CDATA[中等题。]]></description><category></category></item><item><title><![CDATA[Leetcode 221. 最大正方形]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129344986</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129344986</guid><author>qq_43456497</author><pubDate>Sun, 05 Mar 2023 14:09:47 +0800</pubDate><description><![CDATA[中等题。]]></description><category></category></item><item><title><![CDATA[Leetcode 416. 分割等和子集【0-1背包】]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129335383</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129335383</guid><author>qq_43456497</author><pubDate>Sat, 04 Mar 2023 16:11:34 +0800</pubDate><description><![CDATA[中等题， 背包问题。]]></description><category></category></item><item><title><![CDATA[Leetcode 136. 只出现一次的数字]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129303998</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129303998</guid><author>qq_43456497</author><pubDate>Thu, 02 Mar 2023 17:03:50 +0800</pubDate><description><![CDATA[简单题。]]></description><category></category></item><item><title><![CDATA[Leetcode 338. 比特位计数]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129303415</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129303415</guid><author>qq_43456497</author><pubDate>Thu, 02 Mar 2023 16:34:35 +0800</pubDate><description><![CDATA[简单题。]]></description><category></category></item><item><title><![CDATA[Leetcode 152. 乘积最大子数组]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129290199</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129290199</guid><author>qq_43456497</author><pubDate>Wed, 01 Mar 2023 21:37:02 +0800</pubDate><description><![CDATA[中等题。]]></description><category></category></item><item><title><![CDATA[Leetcode 337. 打家劫舍 III]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129289119</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129289119</guid><author>qq_43456497</author><pubDate>Wed, 01 Mar 2023 21:23:08 +0800</pubDate><description><![CDATA[中等题。]]></description><category></category></item><item><title><![CDATA[Leetcode 647. 回文子串【动态规划】]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129288542</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129288542</guid><author>qq_43456497</author><pubDate>Wed, 01 Mar 2023 20:19:10 +0800</pubDate><description><![CDATA[中等题。]]></description><category></category></item><item><title><![CDATA[Leetcode 139. 单词拆分【动态规划】]]></title><link>https://blog.youkuaiyun.com/qq_43456497/article/details/129250881</link><guid>https://blog.youkuaiyun.com/qq_43456497/article/details/129250881</guid><author>qq_43456497</author><pubDate>Tue, 28 Feb 2023 00:49:24 +0800</pubDate><description><![CDATA[中等题。]]></description><category></category></item></channel></rss>