Indefinite kernel network with \(l^q\)-norm regularization (Q1723692)

Summary: We study the asymptotical properties of indefinite kernel network with \(l^q\)-norm regularization. The framework under investigation is different from classical kernel learning. Positive semidefiniteness is not required by the kernel function. By a new step stone technique, without any interior cone condition for input space \(\mathcal{X}\) and \(L_\tau\) condition for the probability measure \(\rho_{\mathcal{X}}\), satisfied error bounds and learning rates are deduced.