[Purpose/significance]This paper aims at finding out the characteristics of author cooperation in computer science. [Method/process] We built the author cooperative network with about 300 minions of papers from 1975 to 2014, and we analyzed the evolution of this network. We built two networks with papers in journals and papers in conferences, and compared the differences of two types of networks. [Result/conclusion] Co-author is a universal phenomenon, and about 80% scientists have more than three cooperators. There are two stages in computer science, the slowly developing stage before 2000 and the rapid stage after 2000. It is found that scientists are more cooperative in conferences than in journals.