Optimizing the CAI (computer assisted instruction) hypertext knowledge base is one of the keys for improving quality of CAI courseware. First, the structure of knowledge points and the mathematical description of their networks are given in the paper. Then the parallel network traveling in “learn first use later” or “learn right use right” manner and the single destination network traveling are given by some layer partitions of network nodes. Some algorithms of reducing the network structure to save storage space and ease operating are introduced. Finally, optimizing the CAI hypertext knowledge base by means of the network properties is summarized.