Online Bayesian Adaptive Sampling for Nonlinear Model: Application to Plant Phenotyping
Gespeichert in:
| Titel: | Online Bayesian Adaptive Sampling for Nonlinear Model: Application to Plant Phenotyping |
|---|---|
| Autoren: | Mercier, Félix, Bouhlel, Nizar, El Ghaziri, Angelina, Rousseau, David |
| Weitere Verfasser: | DUPRE, Olivier |
| Quelle: | 2024 32nd European Signal Processing Conference (EUSIPCO). :2537-2541 |
| Verlagsinformationen: | IEEE, 2024. |
| Publikationsjahr: | 2024 |
| Schlagwörter: | Monte Carlo method, plant phenotyping, [SDV.SA.HORT] Life Sciences [q-bio]/Agricultural sciences/Horticulture, Bayesian approach, compression rate, Markov chain MC, Gompertz function, Gaussian process, Adaptive sampling, [INFO.INFO-IA] Computer Science [cs]/Computer Aided Engineering |
| Beschreibung: | Online adaptive temporal sampling for the monitoring of a nonlinear process of real interest in plant phenomics is developed. We propose to quantify the uncertainty at each sample to be acquired based on prior knowledge (previous samples or statistical priors). Based on this uncertainty, a decision is made whether to take the measurement point by comparing the uncertainty with a tolerance threshold. This approach is illustrated with Monte Carlo, Markov Chain Monte Carlo and Gaussian process modeling. A compression rate of 0.2 is obtained for a large set of genotypes of plants with minimal distortion on the estimated growth curve. This leads to a significant reduction of data storage in such high-throughput plant phenotyping experiments. |
| Publikationsart: | Article Conference object |
| Dateibeschreibung: | application/pdf |
| DOI: | 10.23919/eusipco63174.2024.10715323 |
| Rights: | STM Policy #29 |
| Dokumentencode: | edsair.doi.dedup.....ad488f4d0f9a7daa421446967b49a6dc |
| Datenbank: | OpenAIRE |
Nájsť tento článok vo Web of Science | Abstract: | Online adaptive temporal sampling for the monitoring of a nonlinear process of real interest in plant phenomics is developed. We propose to quantify the uncertainty at each sample to be acquired based on prior knowledge (previous samples or statistical priors). Based on this uncertainty, a decision is made whether to take the measurement point by comparing the uncertainty with a tolerance threshold. This approach is illustrated with Monte Carlo, Markov Chain Monte Carlo and Gaussian process modeling. A compression rate of 0.2 is obtained for a large set of genotypes of plants with minimal distortion on the estimated growth curve. This leads to a significant reduction of data storage in such high-throughput plant phenotyping experiments. |
|---|---|
| DOI: | 10.23919/eusipco63174.2024.10715323 |