A robust adaptive meta-sample generation method for few-shot time series prediction
Abstract The research and exploration of time series prediction (TSP) have attracted much attention recently. Researchers can achieve effective TSP based on the deep learning model and a large amount of data. However, when sufficient high-quality data are not available, the performance of prediction...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2024-12-01
|
| Series: | Complex & Intelligent Systems |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s40747-024-01638-2 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832571150991884288 |
|---|---|
| author | Chao Zhang Defu Jiang Kanghui Jiang Jialin Yang Yan Han Ling Zhu Libo Tao |
| author_facet | Chao Zhang Defu Jiang Kanghui Jiang Jialin Yang Yan Han Ling Zhu Libo Tao |
| author_sort | Chao Zhang |
| collection | DOAJ |
| description | Abstract The research and exploration of time series prediction (TSP) have attracted much attention recently. Researchers can achieve effective TSP based on the deep learning model and a large amount of data. However, when sufficient high-quality data are not available, the performance of prediction models based on deep learning techniques may degrade. Therefore, this paper focuses on few-shot time series prediction (FTSP) and plans to combine meta-learning and generative models to alleviate the problems caused by insufficient training data. When using meta-learning techniques to process FTSP tasks, researchers set the meta-parameter in model-agnostic meta-learning (MAML) as a meta-sample and construct meta-sample generation methods based on advanced generative modeling theory to achieve better uncertainty coding. The existing meta-sample generation methods in FTSP scenes have an inherent limitation: With the increase of the complexity of prediction tasks, samples based on Gaussian distribution may be sensitive to noise and outliers in the meta-learning environment and lack of uncertainty expression, thus affecting the robustness and accuracy of prediction. Therefore, this paper proposes an adaptive sample generation method called JLSG-Diffusion. Based on the Jensen constraint framework and Laplace modeling theory, this method constructs a sample adapter with reasonable adaptive steps and fast convergence for specific tasks. The advantage is to realize fast adaptive convergence of samples to new tasks at lower cost, effectively control the overall generalization error, and improve the robustness and non-Gaussian generalization of sample posterior reasoning. Moreover, the meta sampler of JLSG-Diffusion embeds meta-learning from the implicit probability measure level of Denoising Diffusion Probabilistic Models (DDPM), which makes the meta-sample distribution directly establish a function mapping with the new task and effectively quantifies the uncertainty of spatiotemporal dimension. Experimental results on three real datasets prove the efficiency and effectiveness of JLSG-Diffusion. Compared with the benchmark methods, the prediction model combined with JLSG-Diffusion shows better accuracy. |
| format | Article |
| id | doaj-art-26874243c28248139abcd0da502a036e |
| institution | Kabale University |
| issn | 2199-4536 2198-6053 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | Springer |
| record_format | Article |
| series | Complex & Intelligent Systems |
| spelling | doaj-art-26874243c28248139abcd0da502a036e2025-02-02T12:49:53ZengSpringerComplex & Intelligent Systems2199-45362198-60532024-12-0111112210.1007/s40747-024-01638-2A robust adaptive meta-sample generation method for few-shot time series predictionChao Zhang0Defu Jiang1Kanghui Jiang2Jialin Yang3Yan Han4Ling Zhu5Libo Tao6College of Information Science and Engineering, Hohai UniversityCollege of Information Science and Engineering, Hohai UniversityCollege of Information Science and Engineering, Hohai UniversityCollege of Information Science and Engineering, Hohai UniversityCollege of Information Science and Engineering, Hohai UniversityUnit 93110 of PLAUnit 95806 of PLAAbstract The research and exploration of time series prediction (TSP) have attracted much attention recently. Researchers can achieve effective TSP based on the deep learning model and a large amount of data. However, when sufficient high-quality data are not available, the performance of prediction models based on deep learning techniques may degrade. Therefore, this paper focuses on few-shot time series prediction (FTSP) and plans to combine meta-learning and generative models to alleviate the problems caused by insufficient training data. When using meta-learning techniques to process FTSP tasks, researchers set the meta-parameter in model-agnostic meta-learning (MAML) as a meta-sample and construct meta-sample generation methods based on advanced generative modeling theory to achieve better uncertainty coding. The existing meta-sample generation methods in FTSP scenes have an inherent limitation: With the increase of the complexity of prediction tasks, samples based on Gaussian distribution may be sensitive to noise and outliers in the meta-learning environment and lack of uncertainty expression, thus affecting the robustness and accuracy of prediction. Therefore, this paper proposes an adaptive sample generation method called JLSG-Diffusion. Based on the Jensen constraint framework and Laplace modeling theory, this method constructs a sample adapter with reasonable adaptive steps and fast convergence for specific tasks. The advantage is to realize fast adaptive convergence of samples to new tasks at lower cost, effectively control the overall generalization error, and improve the robustness and non-Gaussian generalization of sample posterior reasoning. Moreover, the meta sampler of JLSG-Diffusion embeds meta-learning from the implicit probability measure level of Denoising Diffusion Probabilistic Models (DDPM), which makes the meta-sample distribution directly establish a function mapping with the new task and effectively quantifies the uncertainty of spatiotemporal dimension. Experimental results on three real datasets prove the efficiency and effectiveness of JLSG-Diffusion. Compared with the benchmark methods, the prediction model combined with JLSG-Diffusion shows better accuracy.https://doi.org/10.1007/s40747-024-01638-2Few-shot time series predictionNon-Gaussian distribution meta-learningRobust adaptive samplingDenoising diffusion probabilistic modelDataset augmentation |
| spellingShingle | Chao Zhang Defu Jiang Kanghui Jiang Jialin Yang Yan Han Ling Zhu Libo Tao A robust adaptive meta-sample generation method for few-shot time series prediction Complex & Intelligent Systems Few-shot time series prediction Non-Gaussian distribution meta-learning Robust adaptive sampling Denoising diffusion probabilistic model Dataset augmentation |
| title | A robust adaptive meta-sample generation method for few-shot time series prediction |
| title_full | A robust adaptive meta-sample generation method for few-shot time series prediction |
| title_fullStr | A robust adaptive meta-sample generation method for few-shot time series prediction |
| title_full_unstemmed | A robust adaptive meta-sample generation method for few-shot time series prediction |
| title_short | A robust adaptive meta-sample generation method for few-shot time series prediction |
| title_sort | robust adaptive meta sample generation method for few shot time series prediction |
| topic | Few-shot time series prediction Non-Gaussian distribution meta-learning Robust adaptive sampling Denoising diffusion probabilistic model Dataset augmentation |
| url | https://doi.org/10.1007/s40747-024-01638-2 |
| work_keys_str_mv | AT chaozhang arobustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT defujiang arobustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT kanghuijiang arobustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT jialinyang arobustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT yanhan arobustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT lingzhu arobustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT libotao arobustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT chaozhang robustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT defujiang robustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT kanghuijiang robustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT jialinyang robustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT yanhan robustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT lingzhu robustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction AT libotao robustadaptivemetasamplegenerationmethodforfewshottimeseriesprediction |