[vc_empty_space][vc_empty_space]
Heteroscedastic gaussian process regression using nearest neighbor point estimates
Robani M.D.a, Palar P.S.a, Zuhal L.R.a
a Institut Teknologi Bandung, Bandung, West Java, 40132, Indonesia
[vc_row][vc_column][vc_row_inner][vc_column_inner][vc_separator css=”.vc_custom_1624529070653{padding-top: 30px !important;padding-bottom: 30px !important;}”][/vc_column_inner][/vc_row_inner][vc_row_inner layout=”boxed”][vc_column_inner width=”3/4″ css=”.vc_custom_1624695412187{border-right-width: 1px !important;border-right-color: #dddddd !important;border-right-style: solid !important;border-radius: 1px !important;}”][vc_empty_space][megatron_heading title=”Abstract” size=”size-sm” text_align=”text-left”][vc_column_text]© 2021, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.When noise corrupts a black-box function, the use of deterministic surrogate models is inadequate since it becomes difficult to distinguish the true function and the noise. Gaussian process is one type of surrogate model that is widely used in practice to handle noisy problems. In several stochastic simulators or physical experiments, the variance of the noise could vary in the input spaces (i.e., heteroscedasticity). This study focuses on the heteroscedastic Gaussian process (HGP) model that utilizes point estimates in learning the noise function without any additional replication. This paper proposes a modified HGP called Nearest Neighbor Point Estimates HGP (NNPEHGP) that uses two GP models similar to the typical HGP configuration. In our proposed model, k-nearest neighbor (kNN) regression is augmented to the procedure to learn the noise functions to overcome the overfitting problem that frequently appears when the number of training samples is small. We test the efficacy of the proposed method on four problems, consisting of two mathematical functions, one real-world dataset, and a heat transfer problem to demonstrate the superiority of the proposed NNPEHGP model in terms of accuracy and stability.[/vc_column_text][vc_empty_space][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][vc_empty_space][megatron_heading title=”Author keywords” size=”size-sm” text_align=”text-left”][vc_column_text][/vc_column_text][vc_empty_space][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][vc_empty_space][megatron_heading title=”Indexed keywords” size=”size-sm” text_align=”text-left”][vc_column_text][/vc_column_text][vc_empty_space][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][vc_empty_space][megatron_heading title=”Funding details” size=”size-sm” text_align=”text-left”][vc_column_text]Pramudita Satria Palar and Lavi Rizki Zuhal were funded in part through the Program Penelitian, Pengabdian Kepada Masyarakat, dan Inovasi (P3MI) 2019-2020 administered by Lembaga Penelitian dan Pengabdian Kepada Masyarakat, Institut Teknologi Bandung, Indonesia.[/vc_column_text][vc_empty_space][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][vc_empty_space][megatron_heading title=”DOI” size=”size-sm” text_align=”text-left”][vc_column_text]https://doi.org/10.2514/6.2021-1589[/vc_column_text][/vc_column_inner][vc_column_inner width=”1/4″][vc_column_text]Widget Plumx[/vc_column_text][/vc_column_inner][/vc_row_inner][/vc_column][/vc_row][vc_row][vc_column][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][/vc_column][/vc_row]