Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Menzen, Clara, Kok, Manon, Batselier, Kim
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2409.03276
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912015632564224
author Menzen, Clara
Kok, Manon
Batselier, Kim
author_facet Menzen, Clara
Kok, Manon
Batselier, Kim
contents The state-of-the-art tensor network Kalman filter lifts the curse of dimensionality for high-dimensional recursive estimation problems. However, the required rounding operation can cause filter divergence due to the loss of positive definiteness of covariance matrices. We solve this issue by developing, for the first time, a tensor network square root Kalman filter, and apply it to high-dimensional online Gaussian process regression. In our experiments, we demonstrate that our method is equivalent to the conventional Kalman filter when choosing a full-rank tensor network. Furthermore, we apply our method to a real-life system identification problem where we estimate $4^{14}$ parameters on a standard laptop. The estimated model outperforms the state-of-the-art tensor network Kalman filter in terms of prediction accuracy and uncertainty quantification.
format Preprint
id arxiv_https___arxiv_org_abs_2409_03276
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Tensor network square root Kalman filter for online Gaussian process regression
Menzen, Clara
Kok, Manon
Batselier, Kim
Machine Learning
The state-of-the-art tensor network Kalman filter lifts the curse of dimensionality for high-dimensional recursive estimation problems. However, the required rounding operation can cause filter divergence due to the loss of positive definiteness of covariance matrices. We solve this issue by developing, for the first time, a tensor network square root Kalman filter, and apply it to high-dimensional online Gaussian process regression. In our experiments, we demonstrate that our method is equivalent to the conventional Kalman filter when choosing a full-rank tensor network. Furthermore, we apply our method to a real-life system identification problem where we estimate $4^{14}$ parameters on a standard laptop. The estimated model outperforms the state-of-the-art tensor network Kalman filter in terms of prediction accuracy and uncertainty quantification.
title Tensor network square root Kalman filter for online Gaussian process regression
topic Machine Learning
url https://arxiv.org/abs/2409.03276