index.html

<!doctype html><html lang=en dir=auto data-theme=auto><head><meta name=generator content="Hugo 0.154.5"><meta charset=utf-8><meta http-equiv=X-UA-Compatible content="IE=edge"><meta name=viewport content="width=device-width,initial-scale=1,shrink-to-fit=no"><meta name=robots content="index, follow"><title>Yan Tang</title><meta name=description content><meta name=author content="Yan Tang"><link rel=canonical href=https://ehehe.cn/><link crossorigin=anonymous href=/assets/css/stylesheet.c9955fba7e0a0fb84b7bcd8f0e496bfbb1c69a88719a87a588a4a271f28c1df4.css integrity="sha256-yZVfun4KD7hLe82PDklr+7HGmohxmoeliKSicfKMHfQ=" rel="preload stylesheet" as=style><link rel=icon href=https://ehehe.cn/assets/images/favicon-16x16.ico><link rel=icon type=image/png sizes=16x16 href=https://ehehe.cn/assets/images/favicon-16x16.ico><link rel=icon type=image/png sizes=32x32 href=https://ehehe.cn/assets/images/favicon-32x32.ico><link rel=apple-touch-icon href=https://ehehe.cn/assets/images/apple-touch-icon.png><link rel=mask-icon href=https://ehehe.cn/safari-pinned-tab.svg><meta name=theme-color content="#2e2e33"><meta name=msapplication-TileColor content="#2e2e33"><link rel=alternate type=application/rss+xml href=https://ehehe.cn/index.xml title=rss><link rel=alternate hreflang=en href=https://ehehe.cn/><noscript><style>#theme-toggle,.top-link{display:none}</style><style>@media(prefers-color-scheme:dark){:root{--theme:rgb(29, 30, 32);--entry:rgb(46, 46, 51);--primary:rgb(218, 218, 219);--secondary:rgb(155, 156, 157);--tertiary:rgb(65, 66, 68);--content:rgb(196, 196, 197);--code-block-bg:rgb(46, 46, 51);--code-bg:rgb(55, 56, 62);--border:rgb(51, 51, 51);color-scheme:dark}.list{background:var(--theme)}.toc{background:var(--entry)}}@media(prefers-color-scheme:light){.list::-webkit-scrollbar-thumb{border-color:var(--code-bg)}}</style></noscript><script>localStorage.getItem("pref-theme")==="dark"?document.querySelector("html").dataset.theme="dark":localStorage.getItem("pref-theme")==="light"?document.querySelector("html").dataset.theme="light":window.matchMedia("(prefers-color-scheme: dark)").matches?document.querySelector("html").dataset.theme="dark":document.querySelector("html").dataset.theme="light"</script><link rel=stylesheet href=https://cdn.jsdelivr.net/npm/katex@0.16.25/dist/katex.min.css integrity=sha384-WcoG4HRXMzYzfCgiyfrySxx90XSl2rxY5mnVY5TwtWE6KLrArNKn0T/mOgNL0Mmi crossorigin=anonymous><script defer src=https://cdn.jsdelivr.net/npm/katex@0.16.25/dist/katex.min.js integrity=sha384-J+9dG2KMoiR9hqcFao0IBLwxt6zpcyN68IgwzsCSkbreXUjmNVRhPFTssqdSGjwQ crossorigin=anonymous></script><meta property="og:url" content="https://ehehe.cn/"><meta property="og:site_name" content="Yan Tang"><meta property="og:title" content="Yan Tang"><meta property="og:locale" content="zh_cn"><meta property="og:type" content="website"><meta name=twitter:card content="summary"><meta name=twitter:title content="Yan Tang"><script type=application/ld+json>{"@context":"https://schema.org","@type":"Organization","name":"Yan Tang","url":"https://ehehe.cn/","description":"","logo":"https://ehehe.cn/assets/images/favicon-16x16.ico","sameAs":["https://github.com/kenanking","mailto:tangyan@tongji.edu.cn","https://x.com/YanTang_TJ"]}</script></head><body class=list id=top><header class=header><nav class=nav><div class=logo><a href=https://ehehe.cn/ accesskey=h title="Yan Tang (Alt + H)">Yan Tang</a><div class=logo-switches><button id=theme-toggle accesskey=t title="(Alt + T)" aria-label="Toggle theme">
<svg id="moon" width="24" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M21 12.79A9 9 0 1111.21 3 7 7 0 0021 12.79z"/></svg>
<svg id="sun" width="24" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><circle cx="12" cy="12" r="5"/><line x1="12" y1="1" x2="12" y2="3"/><line x1="12" y1="21" x2="12" y2="23"/><line x1="4.22" y1="4.22" x2="5.64" y2="5.64"/><line x1="18.36" y1="18.36" x2="19.78" y2="19.78"/><line x1="1" y1="12" x2="3" y2="12"/><line x1="21" y1="12" x2="23" y2="12"/><line x1="4.22" y1="19.78" x2="5.64" y2="18.36"/><line x1="18.36" y1="5.64" x2="19.78" y2="4.22"/></svg></button></div></div><ul id=menu><li><a href=https://ehehe.cn/ title=Posts><span class=active>Posts</span></a></li><li><a href=https://ehehe.cn/about/ title=About><span>About</span></a></li><li><a href=https://ehehe.cn/publications/ title=Publications><span>Publications</span></a></li><li><a href=https://ehehe.cn/archives/ title=Archives><span>Archives</span></a></li><li><a href=https://ehehe.cn/zotwatch/ title=ZotWatch><span>ZotWatch</span></a></li></ul></nav></header><main class=main><article class="first-entry home-info"><div class=home-info-left>👋 Hi, I am a fourth-year PhD student in the College of Surveying and Geo-Informatics at Tongji University, working with <a href=https://celiang.tongji.edu.cn/info/1301/2420.htm>Prof. Shaoming Zhang</a>. I received my B.S. degree from the same institution in 2022.
My research focuses on computer vision and Synthetic Aperture Radar (SAR) target detection. I am particularly interested in developing efficient deep learning methods for SAR image analysis and target recognition, combining computer vision techniques with remote sensing applications.</div><div class=home-info-right><div class=profile_inner><img draggable=false src=https://ehehe.cn/assets/images/avatar.jpeg alt="profile image" title height=150 width=150></div><header class=entry-header><h1>Yan Tang 汤焱</h1></header><footer class=entry-footer><div class=social-icons><a href=https://github.com/kenanking target=_blank rel="noopener noreferrer me" title=Github><svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M9 19c-5 1.5-5-2.5-7-3m14 6v-3.87a3.37 3.37.0 00-.94-2.61c3.14-.35 6.44-1.54 6.44-7A5.44 5.44.0 0020 4.77 5.07 5.07.0 0019.91 1S18.73.65 16 2.48a13.38 13.38.0 00-7 0C6.27.65 5.09 1 5.09 1A5.07 5.07.0 005 4.77 5.44 5.44.0 003.5 8.55c0 5.42 3.3 6.61 6.44 7A3.37 3.37.0 009 18.13V22"/></svg>
</a><a href=mailto:tangyan@tongji.edu.cn target=_blank rel="noopener noreferrer me" title=Email><svg viewBox="0 0 24 21" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M4 4h16c1.1.0 2 .9 2 2v12c0 1.1-.9 2-2 2H4c-1.1.0-2-.9-2-2V6c0-1.1.9-2 2-2z"/><polyline points="22,6 12,13 2,6"/></svg>
</a><a href=https://x.com/YanTang_TJ target=_blank rel="noopener noreferrer me" title=X><svg viewBox="0 0 24 24" fill="currentColor"><path d="M18.244 2.25h3.308l-7.227 8.26 8.502 11.24H16.17l-5.214-6.817L4.99 21.75H1.68l7.73-8.835L1.254 2.25H8.08l4.713 6.231zm-1.161 17.52h1.833L7.084 4.126H5.117z"/></svg></a></div></footer></div></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>LeJEPA：可证明、可扩展的自监督学习新范式</h2></header><div class=entry-content><p>本文记录了 LeJEPA 论文的阅读笔记。</p></div><footer class=entry-footer><span title='2025-11-19 00:00:00 +0000 UTC'>November 19, 2025</span>&nbsp;·&nbsp;<span>1 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to LeJEPA：可证明、可扩展的自监督学习新范式" href=https://ehehe.cn/posts/2025/04-lejepa/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>深入理解 PyTorch SGD 优化器参数</h2></header><div class=entry-content><p>本文通过可视化损失函数和梯度下降路径，深入探讨了 PyTorch SGD 优化器中的各种参数（如动量、权重衰减、Nesterov 动量、动量抑制因子等）对简单线性回归问题的影响，以深入理解神经网络训练中的复杂性。</p></div><footer class=entry-footer><span title='2025-11-05 00:00:00 +0000 UTC'>November 5, 2025</span>&nbsp;·&nbsp;<span>5 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to 深入理解 PyTorch SGD 优化器参数" href=https://ehehe.cn/posts/2025/03-pytorch-sgd-optimizer/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>强化学习玩 Flappy Bird</h2></header><div class=entry-content><p>Flappy Bird 是一款看似简单的移动端游戏，玩家需要控制一只小鸟向前飞行，并穿越一系列障碍物。小鸟只有飞翔和下落两种动作，通过控制小鸟的飞行高度来穿越障碍物。本文记录了我使用 DQN 训练 Flappy Bird 的过程。</p></div><footer class=entry-footer><span title='2025-08-24 20:18:30 +0000 UTC'>August 24, 2025</span>&nbsp;·&nbsp;<span>1 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to 强化学习玩 Flappy Bird" href=https://ehehe.cn/posts/2025/01-flappy-bird-dqn/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>DINOv2 可视化 🦖</h2></header><div class=entry-content><p>介绍如何通过 PCA 方式可视化 DINOv2 模型的图像嵌入表示。</p></div><footer class=entry-footer><span title='2024-11-09 00:00:00 +0000 UTC'>November 9, 2024</span>&nbsp;·&nbsp;<span>6 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to DINOv2 可视化 🦖" href=https://ehehe.cn/posts/2025/02-dino-visualization/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>基于 NeRF 的三维场景生成</h2></header><div class=entry-content><p>经典的三维场景表征方法有体素表示、点云表示和网格表示，这三种表示是直接的、显而易见的，因此归为显式的场景表示类别。这里介绍的 NeRF（Neural Radiance Fields）其实也是一种三维场景表征，但是是一种隐式的场景表示（implicit scene representation），因为它不能像点云、网格、体素一样直接看见一个三维模型，需要将神经表征转换到显示的表征或渲染成可见的图像才可以被看到。</p></div><footer class=entry-footer><span title='2024-02-29 14:20:32 +0000 UTC'>February 29, 2024</span>&nbsp;·&nbsp;<span>2 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to 基于 NeRF 的三维场景生成" href=https://ehehe.cn/posts/2024/01-nerf/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>词嵌入方法（Word2Vec）</h2></header><div class=entry-content><p>本篇文章介绍了Word2Vec方法，该方法通过在给定中心词的情况下预测上下文词的概率来学习单词的分布式表示，从而克服了独热表示的缺点，提高了词汇相似度的表达能力。</p></div><footer class=entry-footer><span title='2024-02-28 23:06:29 +0000 UTC'>February 28, 2024</span>&nbsp;·&nbsp;<span>6 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to 词嵌入方法（Word2Vec）" href=https://ehehe.cn/posts/2024/02-word2vec/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>Swin Transformer：层级式特征图与移动窗口注意力机制</h2></header><div class=entry-content><p>Swin Transformer 是一种创新的 Vision Transformer 架构，通过引入层级式特征图和移动窗口注意力机制，解决了 Vision Transformer 在计算复杂度和多尺度特征提取方面的限制，使其成为计算机视觉任务中高效的骨干网络。</p></div><footer class=entry-footer><span title='2023-08-09 00:00:00 +0000 UTC'>August 9, 2023</span>&nbsp;·&nbsp;<span>7 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to Swin Transformer：层级式特征图与移动窗口注意力机制" href=https://ehehe.cn/posts/2023/02-swin-transformer/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>Vision Transformer —— 图像识别中的 Transformer 架构</h2></header><div class=entry-content><p>本文介绍了 Vision Transformer (ViT) 的核心概念，包括如何将 Transformer 架构应用于图像识别任务，以及与传统 CNNs 的比较。</p></div><footer class=entry-footer><span title='2023-07-26 15:27:24 +0000 UTC'>July 26, 2023</span>&nbsp;·&nbsp;<span>5 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to Vision Transformer —— 图像识别中的 Transformer 架构" href=https://ehehe.cn/posts/2023/01-vit/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>DeepLabv2：基于空洞卷积与 ASPP 的语义图像分割</h2></header><div class=entry-content><p>DeepLabv2 通过空洞卷积和上采样滤波器进行密集特征提取，将在图像分类上训练的网络重新用于语义分割任务。文中进一步提出ASPP以在多个尺度上编码对象以及图像上下文。为了产生语义准确的预测和精细的物体边界分割图，文中还结合了深度卷积神经网络和全连接条件随机场的思想。</p></div><footer class=entry-footer><span title='2023-07-04 00:00:00 +0000 UTC'>July 4, 2023</span>&nbsp;·&nbsp;<span>2 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to DeepLabv2：基于空洞卷积与 ASPP 的语义图像分割" href=https://ehehe.cn/posts/2023/07-deeplabv2/></a></article><article class=post-entry><header class=entry-header><h2 class=entry-hint-parent>机器学习中的爱因斯坦求和（Einsums）</h2></header><div class=entry-content><p>机器学习和统计学中存在大量的线性组合，许多统计学和机器学习中的算法和模型都可以写成矩阵与向量之间的运算。Einsums 是一种表示向量、矩阵和高维数组之间的线性运算的方法。</p></div><footer class=entry-footer><span title='2022-03-20 12:32:44 +0000 UTC'>March 20, 2022</span>&nbsp;·&nbsp;<span>7 min</span>&nbsp;·&nbsp;<span>Yan Tang</span></footer><a class=entry-link aria-label="post link to 机器学习中的爱因斯坦求和（Einsums）" href=https://ehehe.cn/posts/2022/01-einsums/></a></article></main><footer class=footer><span>&copy; 2026 <a href=https://ehehe.cn/>Yan Tang</a></span> ·
<span>Powered by
<a href=https://gohugo.io/ rel="noopener noreferrer" target=_blank>Hugo</a> &
<a href=https://github.com/adityatelange/hugo-PaperMod/ rel=noopener target=_blank>PaperMod</a></span></footer><a href=#top aria-label="go to top" title="Go to Top (Alt + G)" class=top-link id=top-link accesskey=g><svg viewBox="0 0 12 6" fill="currentColor"><path d="M12 6H0l6-6z"/></svg>
</a><script>(function(){function e(){var e,t,n,s=document.getElementsByTagName("code");for(n=0;n<s.length;){if(t=s[n],t.parentNode.tagName!=="PRE"&&t.childElementCount===0&&(e=t.textContent,/^\$[^$]/.test(e)&&/[^$]\$$/.test(e)&&(e=e.replace(/^\$/,"\\(").replace(/\$$/,"\\)"),t.textContent=e),/^\\\((.|\s)+\\\)$/.test(e)||/^\\\[(.|\s)+\\\]$/.test(e)||/^\$(.|\s)+\$$/.test(e)||/^\\begin\{([^}]+)\}(.|\s)+\\end\{[^}]+\}$/.test(e))){t.outerHTML=t.innerHTML;continue}n++}}document.readyState==="loading"?document.addEventListener("DOMContentLoaded",e,{once:!0}):e()})()</script><script>(function(){var n=window.pageYOffset||document.documentElement.scrollTop||0,s=800,t=!1,e=null;function o(){if(t=!1,e||(e=document.getElementById("top-link")),!e)return;var o=window.pageYOffset||document.documentElement.scrollTop||document.body.scrollTop||0,i=o>n;n=o<0?0:o,o>s&&!i?(e.style.visibility="visible",e.style.opacity="1"):(e.style.visibility="hidden",e.style.opacity="0")}window.addEventListener("scroll",function(){t||(window.requestAnimationFrame(o),t=!0)},{passive:!0})})()</script><script>(function(){var t=10;function n(e){if(!e)return[0,20];var n,s,o=window.getComputedStyle(e),t=parseFloat(o.lineHeight);return(!t||isNaN(t))&&(t=20),n=e.getBoundingClientRect(),s=n.height||e.offsetHeight||0,[Math.round(s/t),t]}function e(){var e=document.querySelectorAll(".highlight");if(!e||!e.length)return;Array.prototype.forEach.call(e,function(e){if(e.classList.contains("expanded")||e.classList.contains("collapsible"))return;var s,o,a,r,c,l,i=e.querySelector("pre code");if(!i)return;if(r=i.className||"",r.indexOf("language-mermaid")>=0||e.classList.contains("mermaid"))return;if(a=n(i),c=a[0],l=a[1],c<=t)return;e.classList.add("collapsible"),e.style.setProperty("--code-line-height",l+"px"),e.id||(e.id="code-block-"+Math.random().toString(36).slice(2)),o=document.createElement("div"),o.className="code-expand-wrapper",s=document.createElement("button"),s.type="button",s.className="code-expand-link",s.textContent="Show more",s.setAttribute("aria-expanded","false"),s.setAttribute("aria-controls",e.id),s.addEventListener("click",function(){var n,i,a,o=e.classList.contains("expanded"),r=getComputedStyle(e).getPropertyValue("--code-line-height"),t=parseFloat(r)*(parseFloat(getComputedStyle(e).getPropertyValue("--code-max-lines"))||10);(!isFinite(t)||t<=0)&&(t=10*20),i=e.getBoundingClientRect().height,a=o?t:e.scrollHeight,e.style.maxHeight=i+"px",e.offsetHeight,o?e.classList.remove("expanded"):e.classList.add("expanded"),e.style.maxHeight=a+"px",n=function(t){if(t.propertyName!=="max-height")return;if(e.removeEventListener("transitionend",n),e.style.maxHeight="",e.classList.contains("expanded"))s.textContent="Show less",s.setAttribute("aria-expanded","true");else{s.textContent="Show more",s.setAttribute("aria-expanded","false");try{e.scrollIntoView({behavior:"smooth",block:"nearest"})}catch{}}window.dispatchEvent(new Event("resize"))},e.addEventListener("transitionend",n)}),o.appendChild(s),e.nextSibling?e.parentNode.insertBefore(o,e.nextSibling):e.parentNode.appendChild(o)})}document.readyState==="loading"?document.addEventListener("DOMContentLoaded",e,{once:!0}):e()})()</script><script>let menu=document.getElementById("menu");if(menu){const e=localStorage.getItem("menu-scroll-position");e&&(menu.scrollLeft=parseInt(e,10)),menu.onscroll=function(){localStorage.setItem("menu-scroll-position",menu.scrollLeft)}}document.querySelectorAll('a[href^="#"]').forEach(e=>{e.addEventListener("click",function(e){e.preventDefault();var t=this.getAttribute("href").substr(1);window.matchMedia("(prefers-reduced-motion: reduce)").matches?document.querySelector(`[id='${decodeURIComponent(t)}']`).scrollIntoView():document.querySelector(`[id='${decodeURIComponent(t)}']`).scrollIntoView({behavior:"smooth"}),t==="top"?history.replaceState(null,null," "):history.pushState(null,null,`#${t}`)})})</script><script>var mybutton=document.getElementById("top-link");window.onscroll=function(){document.body.scrollTop>800||document.documentElement.scrollTop>800?(mybutton.style.visibility="visible",mybutton.style.opacity="1"):(mybutton.style.visibility="hidden",mybutton.style.opacity="0")}</script><script>document.getElementById("theme-toggle").addEventListener("click",()=>{const e=document.querySelector("html");e.dataset.theme==="dark"?(e.dataset.theme="light",localStorage.setItem("pref-theme","light")):(e.dataset.theme="dark",localStorage.setItem("pref-theme","dark"))})</script></body></html>