Graduate School and Research Center in Digital Sciences

Deep Gaussian processes for multi-fidelity modeling

Cutajar, Kurt; Pullin, Mark; Damianou, Andreas; Lawrence, Neil; González, Javier

NeurIPS 2018, 32nd Neural Information Processing Systems Conference, 2-8 December, 2018, Montreal, Canada 

Multi-fidelity models are prominently used in various science and engineering applications where cheaply-obtained, but possibly biased and noisy observations must be effectively combined with limited or expensive true data in order to construct reliable models. The notion of applying deep Gaussian processes (DGPs) to this setting has recently shown great promise by capturing complex nonlinear correlations across fidelities. However, the architectures explored thus far are burdened by structural assumptions and constraints which deter such models from performing to the best of their expected capabilities. In this paper we propose a novel approach for DGP multi-fidelity modeling which treats DGP layers as fidelity levels and uses a variational inference scheme to propagate uncertainty across them. In our experiments, we show that this approach makes substantial improvements in quantifying and propagating uncertainty in multi-fidelity set-ups, which in turn improves their effectiveness in decision-making pipelines.

Document Bibtex

Title:Deep Gaussian processes for multi-fidelity modeling
Department:Data Science
Eurecom ref:5755
Bibtex: @inproceedings{EURECOM+5755, year = {2018}, title = {{D}eep {G}aussian processes for multi-fidelity modeling}, author = {{C}utajar, {K}urt and {P}ullin, {M}ark and {D}amianou, {A}ndreas and {L}awrence, {N}eil and {G}onz{\'a}lez, {J}avier}, booktitle = {{N}eur{IPS} 2018, 32nd {N}eural {I}nformation {P}rocessing {S}ystems {C}onference, 2-8 {D}ecember, 2018, {M}ontreal, {C}anada \&\#13;\&\#10;}, address = {{M}ontreal, {CANADA}}, month = {12}, url = {} }
See also: