Application of radial basis functions neutral networks in spectral functions

The reconstruction of spectral function from correlation in Euclidean space is a challenging task. In this paper, we employ the Machine Learning techniques terms radial basis functions networks to reconstruct finite number data. To test our method, first generate one type data using mock by mixing several Breit-Wigner propagators. We found that compared with other traditional methods, TSVD, Tikhonov, and MEM, approach gives continuous unified for both positive definite negative function, which especially useful studying QCD phase transition. Moreover, has considerably better performance low frequency region. This advantages extraction transport coefficients are related zero limit function. With generated through model stress energy tensor, find method precise stable coefficients.