Neural-Based TLS Algorithms

The TLS problem can be solved by using direct and iterative methods. The neural approaches can be considered iterative methods, and they have lower computational complexity compared with other iterative methods, which make them more suitable in real-time application. There are three neural ways of solving TLS problem: (1) One is a neural network for the SVD, which finds the right singular vector associated with the smallest singular value of the argumented input data matrix of by minimizing a cost function with a gradient learning, and a subsequent normalization is needed.