Low-rank attention augmented Gaussian processes for multivariate data analysis

Fuente: Zenodo
Guardado en:
Detalles Bibliográficos
Autor principal: Oyebamiji, Oluwole
Formato: Recurso digital
Publicado: Zenodo 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866902247487569920
author Oyebamiji, Oluwole
author_facet Oyebamiji, Oluwole
contents <p><span>We have developed an efficient low-rank attention-augmented Gaussian processes (LAAGP) </span><span>model that effectively combines accuracy with a reduction in the computational costs associated with transformer attention and Gaussian processes (GP). This model addresses the limitations of standard GP models, such as poor covariance function expressiveness for long-range multivariate forecasting and inadequate </span><span>data representation capacity. LAAGP is a powerful forecasting technique that integrates the transformer </span><span>self-attention mechanism with GP. The framework features a transformer encoder that processes the input </span><span>embeddings to extract essential information, using positional and variable encoding along with relative </span><span>embeddings to enhance attention scores. The GP decoder, known for its flexibility and reliable uncertainty </span><span>estimates, has been adapted to predict the system’s evolution over time. This enhancement enables the model to achieve a balance between computational efficiency, predictive accuracy, and uncertainty quantification, thereby enhancing performance on complex tasks such as</span><span> long-range predictions. Our model has been evaluated </span><span>on several benchmark regression and classification datasets.</span></p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_17385441
institution Zenodo
language
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle Low-rank attention augmented Gaussian processes for multivariate data analysis
Oyebamiji, Oluwole
Low rank approximation
Gaussian process
Transformer
<p><span>We have developed an efficient low-rank attention-augmented Gaussian processes (LAAGP) </span><span>model that effectively combines accuracy with a reduction in the computational costs associated with transformer attention and Gaussian processes (GP). This model addresses the limitations of standard GP models, such as poor covariance function expressiveness for long-range multivariate forecasting and inadequate </span><span>data representation capacity. LAAGP is a powerful forecasting technique that integrates the transformer </span><span>self-attention mechanism with GP. The framework features a transformer encoder that processes the input </span><span>embeddings to extract essential information, using positional and variable encoding along with relative </span><span>embeddings to enhance attention scores. The GP decoder, known for its flexibility and reliable uncertainty </span><span>estimates, has been adapted to predict the system’s evolution over time. This enhancement enables the model to achieve a balance between computational efficiency, predictive accuracy, and uncertainty quantification, thereby enhancing performance on complex tasks such as</span><span> long-range predictions. Our model has been evaluated </span><span>on several benchmark regression and classification datasets.</span></p>
title Low-rank attention augmented Gaussian processes for multivariate data analysis
topic Low rank approximation
Gaussian process
Transformer
url https://doi.org/10.5281/zenodo.17385441