Thanks to the Internet and the World Wide Web, we live in a world of many possibilities we can choose from thousands of movies, millions of books, and billions of web pages. Far exceeding our personal processing capacity, this excessive freedom of choice calls for automated ways to find the relevant information. As a result, the field of information filtering is very active and rich with unanswered challenges. In this short paper, I will give a brief introduction on the design of recommender systems, which recommend objects to users based on the historical records of users’ activities. A diffusion-based recommendation algorithm, as well as two improved algorithms are investigated. Numerical results on a benchmark data set have demonstrated the advantages in algorithmic accuracy.