Site updated: 2024-05-23 12:02:44

htlpt · May 23, 2024 · f758274 · f758274
1 parent 0746d32
commit f758274
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 3 deletions.
diff --git a/2024/05/09/Bayesian-Inference-VAEs/index.html b/2024/05/09/Bayesian-Inference-VAEs/index.html
@@ -89,7 +89,7 @@
 <meta property="og:image" content="https://htplt.github.io/2024/05/09/Bayesian-Inference-VAEs/Reparameterized_Variational_Autoencoder.png">
 <meta property="og:image" content="https://htplt.github.io/2024/05/09/Bayesian-Inference-VAEs/Reparameterization_Trick.png">
 <meta property="article:published_time" content="2024-05-09T05:02:57.000Z">
-<meta property="article:modified_time" content="2024-05-09T09:38:50.547Z">
+<meta property="article:modified_time" content="2024-05-09T14:41:08.034Z">
 <meta property="article:author" content="Silven Huang">
 <meta property="article:tag" content="ML">
 <meta name="twitter:card" content="summary">
@@ -391,7 +391,7 @@ <h2 id="vae-变分自编码器">VAE (变分自编码器)</h2>
 \end{aligned}
 \]</span>
 最大化ELBO，就是最大化观测数据的对数似然，最小化近似后验和精确后验的散度。因此VAE的求解目标就是<span
-class="math inline">\(\theta^\*,\phi^*=\underset{\theta,\phi}{\operatorname*{\arg\max}}L_{\theta,\phi}(x)\)</span>。</p>
+class="math inline">\(\theta^*,\phi^*=\underset{\theta,\phi}{\operatorname*{\arg\max}}L_{\theta,\phi}(x)\)</span>。</p>
 <p>其中<span class="math inline">\(\ln
 p_{\theta}(x|z)\)</span>代表重建损失，可以使用蒙特卡洛采样计算，也可使用<span
 class="math inline">\(-\frac12\|x-D_{\theta}(z)\|_{2}^{2}\)</span>实现。</p>

diff --git a/css/main.css b/css/main.css
@@ -1816,7 +1816,7 @@ pre .javascript .function {
   width: 4px;
   height: 4px;
   border-radius: 50%;
-  background: #8eccff;
+  background: #ff12ff;
 }
 .links-of-blogroll {
   font-size: 13px;