Skip to content

Commit

Permalink
[DOC] add align=center
Browse files Browse the repository at this point in the history
  • Loading branch information
TAAAN committed Jun 12, 2017
1 parent a348a31 commit 975ce56
Showing 1 changed file with 5 additions and 5 deletions.
10 changes: 5 additions & 5 deletions docs/algo/kmeans_on_angel.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,14 +4,14 @@
## 1. 算法介绍

每个样本被划分到距离最近的簇。每个簇所有样本的几何中心为这个簇的簇心,样本到簇心的距离为样本到簇的距离。Kmeans算法一般以迭代的方式训练,如下所示:
![xi](../img/kmeans.png)
其中:![xi](../img/xi.png)代表第i个样本,![ci](../img/ci.png)代表与第i个样本距离最近的簇,![miu_i](../img/miui.png)代表第j个簇的簇心。
每个样本被划分到距离最近的簇。每个簇所有样本的几何中心为这个簇的簇心,样本到簇心的距离为样本到簇的距离。Kmeans算法一般以迭代的方式训练,如下所示:
<div align=center>![xi](../img/kmeans.png)</div>
其中:![xi](../img/xi.png)代表第i个样本,![ci](../img/ci.png)代表与第i个样本距离最近的簇,![miu_i](../img/miu_i.png)代表第j个簇的簇心。


## Mini-batch KMeans
"Web-Scale K-Means Clustering"提出一种在朴素KMeans算法基础上改进的KMeans算法,用mini-batch方法训练,每次迭代选择一个mini-batch的样本更新簇心,如下所示:
![mini_batch_kmeans](../img/mini_batch_kmeans.png)
<div align=center>![mini_batch_kmeans](../img/mini_batch_kmeans.png)</div>

## 2. 分布式实现 on Angel

Expand All @@ -23,7 +23,7 @@ KMeans on Angel以迭代的方式训练,每次训练以上文提到的mini-bat

### 算法逻辑
KMeans on Angel的算法流程如下图所示:
![KMeans_on_Angel](../img/KMeans_on_Angel.png)
<div align=center>![KMeans_on_Angel](../img/KMeans_on_Angel.png)</div>


## 3. 运行 & 性能
Expand Down

0 comments on commit 975ce56

Please sign in to comment.