Residual Multi-Fidelity Neural Network Computing

arXiv:2310.03572v2 Announce Type: replace
Abstract: In this work, we consider the general problem of constructing a neural network surrogate model using multi-fidelity information. Motivated by rigorous error and complexity estimates for ReLU neural networks, given an inexpensive low-fidelity and an expensive high-fidelity computational model, we present a residual multi-fidelity computational framework that formulates the correlation between models as a residual function, a possibly non-linear mapping between 1) the shared input space of the models together with the low-fidelity model output …

abstract arxiv complexity computational computing correlation cs.lg cs.na error fidelity framework general information low math.na network networks neural network neural networks relu residual type work