<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[狂小虎/Anthony Dave的博客 KuangXiaoHu's Blogs]]></title><description><![CDATA[欢迎各位来这里互相学习~Ever dream, never stop!]]></description><link>https://blog.youkuaiyun.com/Davidietop</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; Davidietop]]></copyright><item><title><![CDATA[亲测解决Input and output must have the same number of spatial dimensions]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/149565574</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/149565574</guid><author>Davidietop</author><pubDate>Wed, 23 Jul 2025 11:23:46 +0800</pubDate><description><![CDATA[这个问题是由于两个变量的维度大小不一致导致，解决方法是查验变量读取和处理是否出现问题。
]]></description><category></category></item><item><title><![CDATA[亲测解决无法连接此wifi &Cannot connect to this wifi]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/149565383</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/149565383</guid><author>Davidietop</author><pubDate>Wed, 23 Jul 2025 11:18:48 +0800</pubDate><description><![CDATA[这个问题可以通过先连接手机热点等可用wifi来更新驱动解决，或者通过更换wifi驱动解决。
]]></description><category></category></item><item><title><![CDATA[亲测解决This application failed to start because no Qt platform plugin could be initialized.]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/149565168</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/149565168</guid><author>Davidietop</author><pubDate>Wed, 23 Jul 2025 11:12:29 +0800</pubDate><description><![CDATA[这个问题是由于环境变量改变或者Qt的路径改变导致，解决办法是在环境变量里面重新添加Qt路径。
]]></description><category></category></item><item><title><![CDATA[亲测解决Markdown的数学公式无法在网页渲染Markdown math does not render]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/149565053</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/149565053</guid><author>Davidietop</author><pubDate>Wed, 23 Jul 2025 11:08:15 +0800</pubDate><description><![CDATA[这个问题是由于网页的数学渲染引擎适应性不太好，这里我们可以对公式做一些修改。
]]></description><category></category></item><item><title><![CDATA[亲测解决miktex-console failed to start because no Qt platform plugin could be initialized.]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/149564859</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/149564859</guid><author>Davidietop</author><pubDate>Wed, 23 Jul 2025 11:04:42 +0800</pubDate><description><![CDATA[这个问题可能是miktex版本太低导致，解决方法是升级miktex版本。
]]></description><category></category></item><item><title><![CDATA[06 Deep learning神经网络编程基础 激活函数 --吴恩达]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148535304</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148535304</guid><author>Davidietop</author><pubDate>Mon, 09 Jun 2025 15:06:46 +0800</pubDate><description><![CDATA[实践建议：优先使用ReLU及其变体作为默认选择，在特殊场景（如GAN生成器）可尝试Tanh，注意输出层激活函数需与损失函数匹配。（Google提出）]]></description><category></category></item><item><title><![CDATA[05 Deep learning神经网络编程基础 计算图 --吴恩达]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148528303</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148528303</guid><author>Davidietop</author><pubDate>Mon, 09 Jun 2025 10:47:18 +0800</pubDate><description><![CDATA[反向传播本质是微分运算在计算图上的系统化实现，通过自动微分机制高效计算各参数梯度。现代深度学习框架（如TensorFlow/PyTorch）均基于计算图自动构建正反向传播流程。：输入层→隐藏层→输出层。：链式法则的逆向应用。]]></description><category></category></item><item><title><![CDATA[04 Deep learning神经网络编程基础 梯度下降 --吴恩达]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148475537</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148475537</guid><author>Davidietop</author><pubDate>Fri, 06 Jun 2025 15:31:27 +0800</pubDate><description><![CDATA[对于样本集XXX和参数θ\thetaθ，损失函数LθL(\theta)Lθ衡量预测值与真实值的差异。Lθ−1m∑i1myilog⁡hθxi1−yilog⁡1−hθxiLθ−m1​i1∑m​yiloghθ​xi1−yilog1−hθ​xi)]概念作用数学表达损失函数量化模型预测误差Lθ1m∑liLθm1​∑li​偏导数。]]></description><category></category></item><item><title><![CDATA[亲测解决The scripts pylupdate5.exe, pyrcc5.exe and pyuic5.exe which is not on PATH]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148458937</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148458937</guid><author>Davidietop</author><pubDate>Thu, 05 Jun 2025 21:07:32 +0800</pubDate><description><![CDATA[运行PyQt GUI程序时出现PATH警告提示，原因是缺少PyQt5环境工具。解决方法是通过pip安装pyqt5-tools包，该包包含所需的pylupdate5、pyrcc5和pyuic5等工具。安装命令为"pip install pyqt5-tools"，安装后即可解决路径问题。]]></description><category></category></item><item><title><![CDATA[亲测解决self.transform is not exist]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148458905</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148458905</guid><author>Davidietop</author><pubDate>Thu, 05 Jun 2025 21:04:59 +0800</pubDate><description><![CDATA[摘要 出现该问题是由于torchvision版本过低导致的。解决方法为：先卸载当前版本的torchvision，然后重新安装最新版本。具体操作命令为：先执行&quot;pip uninstall torchvision;卸载旧版本，再执行pip install torchvision&quot;安装最新版本。]]></description><category></category></item><item><title><![CDATA[亲测解决grad can be implicitly created only for scalar outputs]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148458306</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148458306</guid><author>Davidietop</author><pubDate>Thu, 05 Jun 2025 20:32:07 +0800</pubDate><description><![CDATA[PyTorch不支持多loss值直接反向传播，需转化为标量形式。问题出现的原因是多个loss值拼接在一起无法自动求导。解决方法是对输出求平均值后再调用.backward()。示例代码展示了如何将10维输出通过.mean()转化为标量后进行反向传播。这一限制是因为PyTorch的自动微分机制需要明确的标量输出才能计算梯度。]]></description><category></category></item><item><title><![CDATA[03 Deep learning神经网络的编程基础 代价函数（Cost function）--吴恩达]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148453228</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148453228</guid><author>Davidietop</author><pubDate>Thu, 05 Jun 2025 17:15:23 +0800</pubDate><description><![CDATA[摘要：本文系统梳理了深度学习中的损失函数原理与应用，重点分析了逻辑回归的交叉熵损失。通过数学推导展示了从概率表达式到损失函数的转化过程（$J(w,b)=-\frac{1}{m}\sum[y^{(i)}\log\hat{y}^{(i)}+(1-y^{(i)})\log(1-\hat{y}^{(i)})]$），并给出梯度计算公式和Python实现代码。文章还阐明了损失函数的凸性、概率解释等关键特性，以及多分类场景下的softmax扩展。实验部分通过特征矩阵示例（X=[[2,3],[1,4],[5,2]]）演示了]]></description><category></category></item><item><title><![CDATA[02 Deep learning神经网络的编程基础 逻辑回归--吴恩达]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148449208</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148449208</guid><author>Davidietop</author><pubDate>Thu, 05 Jun 2025 15:23:24 +0800</pubDate><description><![CDATA[逻辑回归是一种用于二分类任务的统计方法，通过Sigmoid函数将线性回归结果映射为概率值（0-1）。核心公式为ŷ=σ(wᵀx+b)，其中Sigmoid函数具有可导、输出概率化等特性，便于梯度计算。虽然名称含&quot;回归&quot;，但实际解决分类问题。其Python实现简单，广泛应用于概率预测，但存在梯度消失等局限，在深层网络中逐渐被ReLU替代。典型应用包括神经网络激活和强化学习动作选择。]]></description><category></category></item><item><title><![CDATA[01 Deep learning神经网络的编程基础 二分类--吴恩达]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148406274</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148406274</guid><author>Davidietop</author><pubDate>Thu, 05 Jun 2025 11:39:06 +0800</pubDate><description><![CDATA[二分类任务是监督学习中最基础的问题类型，其目标是将样本划分为两个互斥类别。设样本特征空间为X⊆RnX⊆Rn，输出空间为Y01Y01fX→YfX→Yxx:表示一个输入数据，维度为（nx，1）yy：表示一个输出结果，取值为（0，1）xiyixiyi：表示第i组数据，此处默认为训练数据。（测试数据）X=[x1x2xmx1x2...xm]: 表示所有训练数据的输入值，放在一个（nx，m）的矩阵中。Y=[y1y2y。]]></description><category></category></item><item><title><![CDATA[00 Deep learning神经网络编程基础 回归、拟合、最小二乘法 --吴恩达]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/148400816</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/148400816</guid><author>Davidietop</author><pubDate>Tue, 03 Jun 2025 17:13:59 +0800</pubDate><description><![CDATA[回归分析与拟合方法研究摘要 回归分析和拟合是数据建模中的核心方法。回归分析通过统计方法研究变量间关系，常采用最小二乘法建立数学模型。拟合则用光滑曲线连接数据点，是回归实现的具体数学手段。最小二乘法通过最小化误差平方和求解最优参数，其矩阵形式为β̂=(XᵀX)⁻¹Xᵀy，计算高效且易于理解。本文通过具体示例展示了最小二乘法的推导过程，并提供了Python实现代码。研究还指出最小二乘法存在对异常值敏感、矩阵不可逆等问题。在实际应用中，回归分析通常通过拟合方法实现建模目标，两者在数据科学领域具有紧密的关联性。]]></description><category></category></item><item><title><![CDATA[亲测解决Warning: Accessing non-existent property ‘lineno‘ of module]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/146310528</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/146310528</guid><author>Davidietop</author><pubDate>Mon, 17 Mar 2025 11:20:26 +0800</pubDate><description><![CDATA[这个问题是由于stylus版本过低导致，解决方法是安装最新版本stylus。]]></description><category></category></item><item><title><![CDATA[亲测解决hexo : hexo.ps1 cannot be loaded because running scripts is disabled on this system]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/146310266</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/146310266</guid><author>Davidietop</author><pubDate>Mon, 17 Mar 2025 11:12:21 +0800</pubDate><description><![CDATA[不要用powershell，用原生cmd运行即可。]]></description><category></category></item><item><title><![CDATA[亲测解决detectron2的ValueError: Unsupported type found in checkpoint! model: ＜class ‘dict‘＞]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/146092302</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/146092302</guid><author>Davidietop</author><pubDate>Fri, 07 Mar 2025 12:10:00 +0800</pubDate><description><![CDATA[这个问题是由于预训练模型保存出错所致，解决方法是严格按照规定格式保存。]]></description><category></category></item><item><title><![CDATA[调整solidwords物体透明度]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/146092249</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/146092249</guid><author>Davidietop</author><pubDate>Fri, 07 Mar 2025 12:05:25 +0800</pubDate><description><![CDATA[这个在颜色外观里面，找到透明度修改即可。]]></description><category></category></item><item><title><![CDATA[亲测解决docx文件无法编辑Part of content cannot be edited of docx file]]></title><link>https://blog.youkuaiyun.com/Davidietop/article/details/146091916</link><guid>https://blog.youkuaiyun.com/Davidietop/article/details/146091916</guid><author>Davidietop</author><pubDate>Fri, 07 Mar 2025 11:53:24 +0800</pubDate><description><![CDATA[这个问题是由于部分文字被密码保护。解决方法是转成xml解除限制，再转回docx。]]></description><category></category></item></channel></rss>