This thesis will mainly discuss the sparsed feed-forward neural network, namely, how to determine and delete the redundant neurons and connections in the network. To begin with, the author gives the mathematical definition of feed-forward neural network, and then introduces the partial and topological order to the sparsed algorithm and the learning algorithm of the feed-forward neural network. As a result, the author puts forward the judgement basis of the redundant neurons and connections. According to the self-configuring and self-adjusting tactics, the paper present self-configuring and self-adjusting algorithms which is suitable for feed-forward neural network. The result of the experiment indicates that the above-mentioned sparsed algorithm can not only delete the redundant neurons and connections in the network effectively, but also improve the performance of the network.