AI训练营-数据挖掘-第二周1
需积分: 0 139 浏览量
2022-08-04
12:13:43
上传
评论
收藏 46KB PDF 举报
AI训练营-数据挖掘-第二周
Day 1
1. Learn Graph Data Structure
a. Graph Data Structure Basics (对英文不习惯的同学可以看《算法导论》第22章第一节)
22.1 Representations of graphs.pdf
0.1MB
b.
i. How to store a graph in Python or C++
ii. The limitation of storing a graph in a matrix (two dimensional array)
c. Basic graph algorithm
i. DFS (https://en.wikipedia.org/wiki/Depth-first_search)
ii. BFS (https://en.wikipedia.org/wiki/Breadth-first_search)
d. Bipartite graph (https://en.wikipedia.org/wiki/Bipartite_graph)
i. Coding exercise at home: store a bipartite graph in python
Now, let's think about how to do recommendation by utilizing the graph structure. There are
two major questions we need to answer here: (1) what is the graph? and (2) how to do
recommendation. During all the following days, we will try to give answers to these two
questions and implement our own graph based recommendation algorithms.
What is the graph?
Generally speaking, in the recommendation scenarios, we have two entities: users and items.
Users tend to engage with items and our task is usually defined as
given a user, recommend
top K items to this user.
Obviously, this naturally fits in the graph. Here is an simple bipartite graph illustration.
In the movie recommendation task, we can represent users as the red nodes and movies as
the blue nodes in above graph. If one user watches/likes/comments a movie, we put an edge
between the user (red node) and the movie (blue node). Now, we have the movie
recommendation graph.
Another example would be in the online shopping setting, we view users as the red nodes and
products as the blue nodes. If one user buys a product, we add an edge into the graph. This
will give us the production recommendation graph.
How do we do recommendation by using the graph?
Graphs store all the relations between users and items, which are very valuable for
recommendation. Think about this simple case (feel free to draw this sample graph on paper):
Jack bought Product A and B. Peter bought Product B, C, D. Because of B, C and D is
accessible to Jack and we can suggest C and D to Jack as our recommendation. This is a very
simple toy example to give you a sense how a simple graph based recommendation works.
Along the course, we will learn some basic graph algorithms to make good recommendations.
评论0