图片加上引用说明

youtube_recall/README.cn.md

@@ -27,7 +27,7 @@ Youtube是世界最大的视频网站之一，其推荐系统帮助10亿以上
 下图展示了整个推荐系统框图:
 <p align="center">
 <img src="images/recommendation_system.png" width="500" height="300" hspace='10'/> <br/>
-Figure 1. 推荐系统框图
+Figure 1. 推荐系统框图（出自论文[1]）
 </p>
 
 整个推荐系统有两部分组成: 召回(candidate generation/recall)和排序(ranking)。
@@ -46,7 +46,7 @@ Figure 1. 推荐系统框图
 下图展示了召回模型的网络结构:
 <p align="center">
 <img src="images/model_network.png" width="600" height="500" hspace='10'/> <br/>
-Figure 2. 召回模型网络结构
+Figure 2. 召回模型网络结构（出自论文[1]）
 </p>
 
 - 输入层:用户的浏览序列、搜索序列、人口统计学特征、和其他上下文信息等

youtube_recall/README.md

@@ -9,7 +9,7 @@ YouTube is the world's largest platform for creating, sharing and discovering vi
 The overall structure of the recommendation system is illustrated in Figure 1.
 <p align="center">
 <img src="images/recommendation_system.png" width="500" height="300" hspace='10'/> <br/>
-Figure 1. Recommendation system architecture
+Figure 1. Recommendation system architecture[1]
 </p>
 
 The system is comprised of two neural networks: one for candidate generation and one for ranking.
@@ -28,7 +28,7 @@ where ![](https://www.zhihu.com/equation?tex=%5Cmathbf%7Bu%7D%5Cin%20%5Cmathbb%7
 Figure 2 shows the general network architecture of candidate generation model:
 <p align="center">
 <img src="images/model_network.png" width="600" height="500" hspace='10'/> <br/>
-Figure 2. Candidate generation model architecture
+Figure 2. Candidate generation model architecture[1]
 </p>
 
 - Input layer: A user's watch history is represented by a variable-length sequence of sparse video IDs, and search history is similarly represented by a variable-length sequence of search tokens.

youtube_recall/test.py

-#!/usr/bin/env python
-# -*- coding: utf-8 -*-
-########################################################################
-# 
-# Copyright (c) 2018 Baidu.com, Inc. All Rights Reserved
-# 
-########################################################################
-"""
-File: test.py
-Author: baidu(baidu@baidu.com)
-Date: 2018/01/12 11:41:37
-"""
-import cPickle
-#with open("./output/item_freq.pkl") as f:
-with open("./data/nid_dict.pkl") as f:
-    item_freq = cPickle.load(f)
-print item_freq