<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[Thinkobj的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/Thinkobj</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; Thinkobj]]></copyright><item><title><![CDATA[《百面机器学习》学习笔记之经典算法(支持向量机SVM、逻辑回归LR、决策树DT)]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/129947840</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/129947840</guid><author>Thinkobj</author><pubDate>Tue, 04 Apr 2023 11:01:22 +0800</pubDate><description><![CDATA[支持向量机（Support Vector Machine, SVM）是监督学习方法。将低维的数据映射到高维的核映射空间，寻找“分类超平面”将不同类别的数据区分开。]]></description><category></category></item><item><title><![CDATA[WSL测试笔记]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128387378</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128387378</guid><author>Thinkobj</author><pubDate>Tue, 20 Dec 2022 17:06:42 +0800</pubDate><description><![CDATA[WSL]]></description><category></category></item><item><title><![CDATA[SNPE-Benchmark]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128387271</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128387271</guid><author>Thinkobj</author><pubDate>Tue, 20 Dec 2022 16:58:19 +0800</pubDate><description><![CDATA[snpe-benchmark]]></description><category></category></item><item><title><![CDATA[MMCV核心组件知识整理]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128386140</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128386140</guid><author>Thinkobj</author><pubDate>Tue, 20 Dec 2022 15:57:03 +0800</pubDate><description><![CDATA[MMCV核心组件知识整理]]></description><category></category></item><item><title><![CDATA[MM Segmentation 训练自己的数据集]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128386042</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128386042</guid><author>Thinkobj</author><pubDate>Tue, 20 Dec 2022 15:54:01 +0800</pubDate><description><![CDATA[MM Segmentation 训练自己的数据集
]]></description><category></category></item><item><title><![CDATA[MMSegmentation官方使用文档记录]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128384809</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128384809</guid><author>Thinkobj</author><pubDate>Tue, 20 Dec 2022 14:47:07 +0800</pubDate><description><![CDATA[mmsegmentation官方使用文档记录]]></description><category></category></item><item><title><![CDATA[MMsegmentation安装教程]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128384657</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128384657</guid><author>Thinkobj</author><pubDate>Tue, 20 Dec 2022 14:42:33 +0800</pubDate><description><![CDATA[mmsegmentation第一次使用记录]]></description><category></category></item><item><title><![CDATA[Git常用命令]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128374031</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128374031</guid><author>Thinkobj</author><pubDate>Mon, 19 Dec 2022 16:11:57 +0800</pubDate><description><![CDATA[git常用命令]]></description><category></category></item><item><title><![CDATA[【CVPR2022】Boosting Robustness of Image Matting with Context Assembling and Strong Data Augmentation]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128330406</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128330406</guid><author>Thinkobj</author><pubDate>Thu, 15 Dec 2022 15:14:59 +0800</pubDate><description><![CDATA[Boosting Robustness of Image Matting with Context Assembling and Strong Data Augmentation]]></description><category></category></item><item><title><![CDATA[【CVPR2022】MatteFormer: Transformer-Based Image Matting via Prior-Tokens]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128209388</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128209388</guid><author>Thinkobj</author><pubDate>Tue, 06 Dec 2022 20:11:48 +0800</pubDate><description><![CDATA[【CVPR2022】 MatteFormer: Transformer-Based Image Matting via Prior-Tokens]]></description><category></category></item><item><title><![CDATA[【ECCV2022】论文阅读笔记TransMatting: Enhancing Transparent Objects Matting with Transformers]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128021324</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128021324</guid><author>Thinkobj</author><pubDate>Thu, 24 Nov 2022 16:28:08 +0800</pubDate><description><![CDATA[ECCV2022 TransMatting: Enhancing Transparent Objects Matting with Transformers]]></description><category></category></item><item><title><![CDATA[【ACCV2022】论文阅读笔记Lightweight Alpha Matting Network Using Distillation-Based Channel Pruning]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/128001320</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/128001320</guid><author>Thinkobj</author><pubDate>Wed, 23 Nov 2022 15:31:54 +0800</pubDate><description><![CDATA[ECCV2022论文]]></description><category></category></item><item><title><![CDATA[【CVPR2022】论文阅读笔记Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/127854298</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/127854298</guid><author>Thinkobj</author><pubDate>Mon, 14 Nov 2022 20:05:08 +0800</pubDate><description><![CDATA[Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels ]]></description><category></category></item><item><title><![CDATA[mmsegmentation使用.csv文件读取数据]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/127792946</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/127792946</guid><author>Thinkobj</author><pubDate>Thu, 10 Nov 2022 17:39:23 +0800</pubDate><description><![CDATA[mmsegmentation使用csv文件读取数据]]></description><category></category></item><item><title><![CDATA[mmsegmentation训练HRF数据集]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/127571058</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/127571058</guid><author>Thinkobj</author><pubDate>Fri, 28 Oct 2022 14:37:09 +0800</pubDate><description><![CDATA[使用mmsegmentation框架训练和测试HRF数据集]]></description><category></category></item><item><title><![CDATA[Cmake学习之find_package]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/126116750</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/126116750</guid><author>Thinkobj</author><pubDate>Tue, 02 Aug 2022 10:15:34 +0800</pubDate><description><![CDATA[find_package学习]]></description><category></category></item><item><title><![CDATA[Cmake学习之cmake入门]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/126116394</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/126116394</guid><author>Thinkobj</author><pubDate>Tue, 02 Aug 2022 10:08:04 +0800</pubDate><description><![CDATA[cmake入门]]></description><category></category></item><item><title><![CDATA[《百面机器学习》学习笔试之模型评估（第2章）]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/121585547</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/121585547</guid><author>Thinkobj</author><pubDate>Sat, 27 Nov 2021 23:30:53 +0800</pubDate><description><![CDATA[01 评估指标的局限性
准确率（Accuracy）的局限性
Accuracy=ncorrectntotal
Accuracy = \frac{{{n_{correct}}}}{{{n_{total}}}}
Accuracy=ntotal​ncorrect​​
当负样本占99%时，分类器把所有样本都预测为负样本也可以获得99%的准确率。所以当不同类别的样本比例非常不均衡时，占比大的类别往往成为影响准确率的最主要因素。
为了解决这个问题，可以使用更为有效地平均准确率（每个类别下的样本准确率的算术平均）作为模型评]]></description><category></category></item><item><title><![CDATA[《百面机器学习》学习笔试之特征工程（第1章）]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/121346902</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/121346902</guid><author>Thinkobj</author><pubDate>Tue, 16 Nov 2021 00:30:43 +0800</pubDate><description><![CDATA[01、特征归一化（为什么需要对数值型的特征归一化？）
对数值类型的特征归一化可以将所有的特征都统一到一个大致相同的数值区间内。方法：
1）线性函数归一化（Min-Max scaling）
Xnorm=X−Xmin⁡Xmax⁡−Xmin⁡
{X_{norm}} = \frac{{X - {X_{\min }}}}{{{X_{\max }} - {X_{\min }}}}
Xnorm​=Xmax​−Xmin​X−Xmin​​
2）零均值归一化（Z-score Normalization）
z=x−μσ
z = ]]></description><category></category></item><item><title><![CDATA[天池新闻推荐入门赛之排序模型+模型融合]]></title><link>https://blog.youkuaiyun.com/Thinkobj/article/details/110789219</link><guid>https://blog.youkuaiyun.com/Thinkobj/article/details/110789219</guid><author>Thinkobj</author><pubDate>Sun, 06 Dec 2020 23:22:02 +0800</pubDate><description><![CDATA[排序模型
通过召回的操作， 我们已经进行了问题规模的缩减， 对于每个用户， 选择出了N篇文章作为了候选集，并基于召回的候选集构建了与用户历史相关的特征，以及用户本身的属性特征，文章本省的属性特征，以及用户与文章之间的特征，下面就是使用机器学习模型来对构造好的特征进行学习，然后对测试集进行预测，得到测试集中的每个候选集用户点击的概率，返回点击概率最大的topk个文章，作为最终的结果。
排序阶段选择了三个比较有代表性的排序模型，它们分别是：
LGB的排序模型
LGB的分类模型
深度学习的分类模型DIN
得到了最]]></description><category></category></item></channel></rss>