8th International Conference on
Computational Harmonic Analysis

September 12-16, 2022

Ingolstadt, Germany

"Local approximation by eignets"

Mao, Tong

An eignet is an analogue of the classical radial basis function networks, developed for manifold learning. In the spirit of distributed learning on manifolds, we design a wavelet-like representation for local approximation of functions on data-defined spaces using eignets, so that each term of this representation is an eignet where a subnetwork can characterize the local smoothness of the target function.

« back