fix: http to https
This commit is contained in:
@@ -9,7 +9,7 @@
|
||||
**树结构**
|
||||
|
||||
<div align=center>
|
||||
<img src="http://ryluo.oss-cn-chengdu.aliyuncs.com/图片image-20220420213149324.png" alt="image-20210308142624189" style="zoom: 15%;" />
|
||||
<img src="https://ryluo.oss-cn-chengdu.aliyuncs.com/图片image-20220420213149324.png" alt="image-20210308142624189" style="zoom: 15%;" />
|
||||
</div>
|
||||
|
||||
如上图,树中的每一个叶子节点对应一个商品item,非叶子结点表示的是item的集合**(这里的树不限于二叉树)**。这种层次化结构体现了粒度从粗到细的item架构。
|
||||
@@ -17,7 +17,7 @@
|
||||
**整体结构**
|
||||
|
||||
<div align=center>
|
||||
<img src="http://ryluo.oss-cn-chengdu.aliyuncs.com/图片image-20220420200433442.png" alt="image-20210308142624189" style="zoom: 70%;" />
|
||||
<img src="https://ryluo.oss-cn-chengdu.aliyuncs.com/图片image-20220420200433442.png" alt="image-20210308142624189" style="zoom: 70%;" />
|
||||
</div>
|
||||
|
||||
# 算法详解
|
||||
@@ -34,7 +34,7 @@
|
||||
2. 对兴趣进行建模
|
||||
|
||||
<div align=center>
|
||||
<img src="http://ryluo.oss-cn-chengdu.aliyuncs.com/图片image-20220420214040264.png" alt="image-20210308142624189" style="zoom: 50%;" />
|
||||
<img src="https://ryluo.oss-cn-chengdu.aliyuncs.com/图片image-20220420214040264.png" alt="image-20210308142624189" style="zoom: 50%;" />
|
||||
</div>
|
||||
|
||||
如上图,用户对叶子层item6感兴趣,可以认为它的兴趣是1,同层别的候选节点的兴趣为0,顺着着绿色线路上去的节点都标记为1,路线上的同层别的候选节点都标记为0。这样的操作就可以根据1和0构建用于每一层的正负样本。
|
||||
@@ -44,7 +44,7 @@
|
||||
3. 训练过程
|
||||
|
||||
<div align=center>
|
||||
<img src="http://ryluo.oss-cn-chengdu.aliyuncs.com/图片image-20220420220831318.png" alt="image-20210308142624189" style="zoom: 15%;" />
|
||||
<img src="https://ryluo.oss-cn-chengdu.aliyuncs.com/图片image-20220420220831318.png" alt="image-20210308142624189" style="zoom: 15%;" />
|
||||
</div>
|
||||
|
||||
整体联合训练的方式如下:
|
||||
|
||||
Reference in New Issue
Block a user