TY  - RPRT
TI  - On structure-exploiting trust-region regularized nonlinear least squares algorithms for neural-network learning.
AU  - Eiji Mizutani
AU  - James W Demmel
DO  - 10.1016/s0893-6080(03)00085-6
UR  - https://pubmed.ncbi.nlm.nih.gov/12850030/
ID  - 12850030
ER  -