Skip to content

Commit

Permalink
terminology and typo fix
Browse files Browse the repository at this point in the history
  • Loading branch information
astonzhang committed Mar 16, 2018
1 parent 34c2230 commit 89bb65f
Show file tree
Hide file tree
Showing 2 changed files with 41 additions and 93 deletions.
132 changes: 40 additions & 92 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ build服务器在 http://ci.mxnet.io 。这台服务器有两块Nvidia M60。

在本地可以如下build html(需要GPU支持)

```bash
```{.python .input}
conda env update -f build/build.yml
source activate gluon_zh_docs
make html
Expand All @@ -24,7 +24,7 @@ make html

如果没有改动notebook里面源代码,所以不想执行notebook,可以使用

```
```{.python .input}
make html EVAL=0
```

Expand All @@ -34,11 +34,11 @@ make html EVAL=0

编译pdf版本需要xelatex,和思源字体。在Ubuntu可以这样安装。

```bash
```{.python .input}
sudo apt-get install texlive-full
```

```bash
```{.python .input}
wget https://github.com/adobe-fonts/source-han-sans/raw/release/OTF/SourceHanSansHWSC.zip
wget https://github.com/adobe-fonts/source-han-serif/raw/release/OTF/SourceHanSerifSC_EL-M.zip
unzip SourceHanSansHWSC.zip
Expand All @@ -51,131 +51,79 @@ sudo fc-cache -f -v

然后可以编译了。

```bash
```{.python .input}
make pdf
```

## 英汉术语对照

## Terminology 中英术语对照表

action, 动作

adversarial learning, 对抗学习
accuracy,准确率

agent, 智能体
baseline,基准

attribute space, 属性空间
batch,批量

attribute value, 属性值
bias,偏差

attribute, 属性
binary classification,二元分类

binary classification, 二分类
bucketing,分桶

classification, 分类
class,类

cluster, 簇
classification,分类

clustering, 聚类
collaborative filtering,协同过滤

confidence, 确信度
cost,成本

contextual bandit problem, 情境式赌博机问题
cross-entropy,交叉熵

covariate shift, 协变量转移
data set,数据集

credit assignment problem, 信用分配问题
decision boundary,决策边界

cross-entropy, 交叉熵
dense,稠密

data set, 数据集
dense layer,稠密层

dimensionality, 维数
dropout,丢弃法

distribution, 分布
empirical risk minimization,经验风险最小化

reinforcement learning, 强化学习
epoch,周期

example, 样例

feature vector, 特征向量
example,样本

feature, 特征

generalization, 泛化

generative adversarial networks, 生成对抗网络

ground-truth, 真相、真实

hypothesis, 假设

independent and identically distributed(i.i.d), 独立同分布

instance, 示例

label space, 标注空间

label, 标注

learing algorithm, 学习算法

learned model, 学得模型

learner, 学习器

learning, 学习

machine translation, 机器翻译

Markov Decision Process, 马尔可夫决策过程

model, 模型

multi-armed bandit problem, 多臂赌博机问题

multi-class classification, 多分类

negative class, 反类

offline learning, 离线学习

positive class, 正类

prediction, 预测

principal component analysis, 主成分分析

regression, 回归

reinforcement learning, 强化学习

representation learning, 表征学习
fully connected layer,全连接层

sample space, 样本空间
hidden layer,隐藏层

sample, 样本
hidden variable,隐藏变量

sepecilization, 特化
generalization,泛化

sequence learning, 序列学习
hypothesis,假设

subspace estimation, 子空间估计
independent and identically distributed(i.i.d),独立同分布

supervised learning, 监督学习
instance,实例

testing sample, 测试样本
label,标签

testing, 测试
logistic regression,逻辑回归

time step, 时间步长
mean squared error,均方误差

training data, 训练数据
metric,指标

training sample, 训练样本
normalization,归一化

training set, 训练集
optimizer,优化器

training, 训练
perplexity,困惑度

unsupervised learning, 无监督学习
pipeline,流水线
2 changes: 1 addition & 1 deletion chapter_natural-language-processing/seq2seq-attention.md
Original file line number Diff line number Diff line change
Expand Up @@ -103,7 +103,7 @@ $$\mathbf{s}_{t^\prime} = g(\mathbf{y}_{t^\prime-1}, \mathbf{c}_{t^\prime}, \mat
$$\mathbf{c}_{t^\prime} = \sum_{t=1}^T \alpha_{t^\prime t} \mathbf{h}_t$$


也就是说,给定解码器的当前时刻$t^\prime$,我们需要对解码器中不同时刻的隐含层变量求加权平均。而权值也称注意力权重。它的计算公式是
也就是说,给定解码器的当前时刻$t^\prime$,我们需要对编码器中不同时刻$t$的隐含层变量求加权平均。而权值也称注意力权重。它的计算公式是

$$\alpha_{t^\prime t} = \frac{\exp(e_{t^\prime t})}{ \sum_{k=1}^T \exp(e_{t^\prime k}) } $$

Expand Down

0 comments on commit 89bb65f

Please sign in to comment.