. Conclusion
In this paper, two neural network approaches are presented for
signal reconstruction of compressed sensing. First, by introducing
implicit variables, we convert the original signal reconstruction
model into a quadratic programming problem. Second, we construct
a nonnegative energy function based on a class of generalized
Fischer–Burmeister complementarity functions. Then,the twoneural
network models for the signal reconstruction of compressed
sensing are established. The two neural networks can be implemented
using integrated circuits. Last, there are some potential
applications of the compressed sensing presented. The neural network
approach can make signal reconstruction of compressed
sensing in real-time.