https://fupcode.com/blog/tags/dl/南冥鹏影Hugo 0.168.0 & FixIt v1.0.0-beta.4zh-CNThu, 17 Sep 2026 17:00:00 +0800https://fupcode.com/blog/posts/research-notes/paper-review-redial/Thu, 17 Sep 2026 17:00:00 +0800https://fupcode.com/blog/posts/research-notes/paper-review-redial/科研笔记通过单点突变观察 embedding 的变化，直接提取 RNA 语言模型中的结构耦合信号，并用逐层分析和消融检验模型冗余。https://fupcode.com/blog/posts/deep-learning/lora/Thu, 20 Nov 2025 19:05:47 +0800https://fupcode.com/blog/posts/deep-learning/lora/深度学习<p>在深度学习微调大模型时，<strong>LoRA（Low-Rank Adaptation）</strong> 提供了一种高效、轻量的微调方法。本文讲述了LoRA的原理，同时结合一个小型 MNIST 实验做了验证。</p>https://fupcode.com/blog/posts/research-notes/paper-review-rhofold/Sat, 15 Nov 2025 17:07:04 +0800https://fupcode.com/blog/posts/research-notes/paper-review-rhofold/科研笔记<h2 class="heading-element" id="模型架构"><span>模型架构</span> <a href="#%e6%a8%a1%e5%9e%8b%e6%9e%b6%e6%9e%84" class="heading-mark"><svg class="octicon octicon-link" viewBox="0 0 16 16" version="1.1" width="16" height="16" aria-hidden="true"><path d="m7.775 3.275 1.25-1.25a3.5 3.5 0 1 1 4.95 4.95l-2.5 2.5a3.5 3.5 0 0 1-4.95 0 .751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018 1.998 1.998 0 0 0 2.83 0l2.5-2.5a2.002 2.002 0 0 0-2.83-2.83l-1.25 1.25a.751.751 0 0 1-1.042-.018.751.751 0 0 1-.018-1.042Zm-4.69 9.64a1.998 1.998 0 0 0 2.83 0l1.25-1.25a.751.751 0 0 1 1.042.018.751.751 0 0 1 .018 1.042l-1.25 1.25a3.5 3.5 0 1 1-4.95-4.95l2.5-2.5a3.5 3.5 0 0 1 4.95 0 .751.751 0 0 1-.018 1.042.751.751 0 0 1-1.042.018 1.998 1.998 0 0 0-2.83 0l-2.5 2.5a1.998 1.998 0 0 0 0 2.83Z"></path></svg></a> </h2><p>RhoFold+是一个全可微分的端到端框架，从RNA序列输入直接输出全原子3D结构。架构分为几个关键模块：</p>https://fupcode.com/blog/posts/deep-learning/cnn-params/Fri, 07 Nov 2025 00:40:31 +0800https://fupcode.com/blog/posts/deep-learning/cnn-params/深度学习<h2 class="heading-element" id="卷积层convolutional-layer-conv-k-n"><span>卷积层（Convolutional Layer, CONV-K-N）</span> <a href="#%e5%8d%b7%e7%a7%af%e5%b1%82convolutional-layer-conv-k-n" class="heading-mark"><svg class="octicon octicon-link" viewBox="0 0 16 16" version="1.1" width="16" height="16" aria-hidden="true"><path d="m7.775 3.275 1.25-1.25a3.5 3.5 0 1 1 4.95 4.95l-2.5 2.5a3.5 3.5 0 0 1-4.95 0 .751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018 1.998 1.998 0 0 0 2.83 0l2.5-2.5a2.002 2.002 0 0 0-2.83-2.83l-1.25 1.25a.751.751 0 0 1-1.042-.018.751.751 0 0 1-.018-1.042Zm-4.69 9.64a1.998 1.998 0 0 0 2.83 0l1.25-1.25a.751.751 0 0 1 1.042.018.751.751 0 0 1 .018 1.042l-1.25 1.25a3.5 3.5 0 1 1-4.95-4.95l2.5-2.5a3.5 3.5 0 0 1 4.95 0 .751.751 0 0 1-.018 1.042.751.751 0 0 1-1.042.018 1.998 1.998 0 0 0-2.83 0l-2.5 2.5a1.998 1.998 0 0 0 0 2.83Z"></path></svg></a> </h2><p><strong>功能</strong>：提取局部特征。</p>https://fupcode.com/blog/posts/deep-learning/mlp-manual/Fri, 07 Nov 2025 00:34:24 +0800https://fupcode.com/blog/posts/deep-learning/mlp-manual/深度学习<h2 class="heading-element" id="一数据集的下载和处理"><span>一、数据集的下载和处理</span> <a href="#%e4%b8%80%e6%95%b0%e6%8d%ae%e9%9b%86%e7%9a%84%e4%b8%8b%e8%bd%bd%e5%92%8c%e5%a4%84%e7%90%86" class="heading-mark"><svg class="octicon octicon-link" viewBox="0 0 16 16" version="1.1" width="16" height="16" aria-hidden="true"><path d="m7.775 3.275 1.25-1.25a3.5 3.5 0 1 1 4.95 4.95l-2.5 2.5a3.5 3.5 0 0 1-4.95 0 .751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018 1.998 1.998 0 0 0 2.83 0l2.5-2.5a2.002 2.002 0 0 0-2.83-2.83l-1.25 1.25a.751.751 0 0 1-1.042-.018.751.751 0 0 1-.018-1.042Zm-4.69 9.64a1.998 1.998 0 0 0 2.83 0l1.25-1.25a.751.751 0 0 1 1.042.018.751.751 0 0 1 .018 1.042l-1.25 1.25a3.5 3.5 0 1 1-4.95-4.95l2.5-2.5a3.5 3.5 0 0 1 4.95 0 .751.751 0 0 1-.018 1.042.751.751 0 0 1-1.042.018 1.998 1.998 0 0 0-2.83 0l-2.5 2.5a1.998 1.998 0 0 0 0 2.83Z"></path></svg></a> </h2><h3 class="heading-element" id="下载数据"><span>下载数据</span> <a href="#%e4%b8%8b%e8%bd%bd%e6%95%b0%e6%8d%ae" class="heading-mark"><svg class="octicon octicon-link" viewBox="0 0 16 16" version="1.1" width="16" height="16" aria-hidden="true"><path d="m7.775 3.275 1.25-1.25a3.5 3.5 0 1 1 4.95 4.95l-2.5 2.5a3.5 3.5 0 0 1-4.95 0 .751.751 0 0 1 .018-1.042.751.751 0 0 1 1.042-.018 1.998 1.998 0 0 0 2.83 0l2.5-2.5a2.002 2.002 0 0 0-2.83-2.83l-1.25 1.25a.751.751 0 0 1-1.042-.018.751.751 0 0 1-.018-1.042Zm-4.69 9.64a1.998 1.998 0 0 0 2.83 0l1.25-1.25a.751.751 0 0 1 1.042.018.751.751 0 0 1 .018 1.042l-1.25 1.25a3.5 3.5 0 1 1-4.95-4.95l2.5-2.5a3.5 3.5 0 0 1 4.95 0 .751.751 0 0 1-.018 1.042.751.751 0 0 1-1.042.018 1.998 1.998 0 0 0-2.83 0l-2.5 2.5a1.998 1.998 0 0 0 0 2.83Z"></path></svg></a> </h3><p>直接调用官方库</p>