VAE
VAE
假定所有数据\(\mathbf{x}^{(1)},\ldots,\mathbf{x}^{(N)}\)都是i.i.d, \[
\log p_{\theta}(\mathbf{x}^{(1)},\ldots,\mathbf{x}^{(N)}) = \sum^N_{i=1}\log p_{\t
...
Min Max & Max Min
发表于
Min Max & Max Min
\[
\begin{align}
\min_y f(x,y) &\le f(x,y),\, \forall x,y\\\
\max_x \min_y f(x,y) &\le \max_x f(x,y),\, \forall y \
...
Improved Training of Wasserstein GANs
发表于
Improved Training of Wasserstein GANs
设G是生成网络(generator)，D是判别网络(critic)，原始GAN的训练目标是 \[
\min_G\max_D\ \mathbb{E}_{\mathbf{x}\sim\mathbb{P}_r}[\log(D(
...
各种Normalization
发表于
Batch/Layer/Instance/Weight Normalization
从使用最广泛的batch norm说起。
Batch Normalization
batch norm是为了解决网络中各个层的输入数据的分布不稳定的问题( (internal) covariate shift)
...
NCE and Negative Sampling
发表于
在语言模型中，给定上下文c，要预测单词w: \[
\begin{align}
p_{\theta}(w|c) = \frac{u_{\theta}(w,c)}{\sum_{w'\in V}u_{\theta}(w',c)} = \frac{u_{\theta}(w,c)}{Z_{
...
Queue, Coordinator and Producer in Tensorflow
发表于
直接用代码 举例说明。
先准备数据： 12345678910import tensorflow as tfimport numpy as npnum_examples = 10num_features = 2data = np.reshape(np.arange(num_examples*num_
...
First Post Test
发表于
This is a testing.
I hope it will work.
今夕复何夕，共此灯烛光
\[E=mc^2\]
\[\hat{H} \lvert\psi\rangle = i\hbar\frac{\partial}{\partial t}\lvert\psi\r
...