通过分解自适应提高时间序列预测的可迁移性
Improving the Transferability of Time Series Forecasting with Decomposition Adaptation
AI总结：
针对时间序列预测中数据稀缺限制模型性能的问题，提出SeDAN迁移架构，通过Implicit Contrastive Decomposition将特征分解为季节和趋势成分并分别进行自适应对齐，在五个基准数据集上验证了其高效稳定的跨域知识迁移能力。
摘要：
由于有效的模式挖掘和特征表示，基于深度学习的神经预测模型取得了巨大进展。有效学习的前提是收集足够的数据。然而，在时间序列预测中，很难获得足够的数据，这限制了神经预测模型的性能。为了缓解数据稀缺的限制，我们设计了 Sequence Decomposition Adaptation Network (SeDAN)，这是一种新颖的迁移架构，通过从跨域数据集中对齐可迁移知识来提高目标域的预测性能。重新思考时间序列数据中特征的可迁移性，我们提出 Implicit Contrastive Decomposition，将原始特征分解为包括季节和趋势特征在内的组件，这些组件更容易迁移。然后我们针对不同域中分解后的特征设计了相应的自适应方法。具体而言，对于季节特征，我们执行联合分布自适应；对于趋势特征，我们设计了 Optimal Local Adaptation。我们在五个基准数据集上进行了广泛的多变量时间序列预测实验。结果表明我们的 SeDAN 的有效性。它可以提供更高效和稳定的迁移。
英文摘要：
Due to effective pattern mining and feature representation, neural forecasting models based on deep learning have achieved great progress. The premise of effective learning is to collect sufficient data. However, in time series forecasting, it is difficult to obtain enough data, which limits the performance of neural forecasting models. To alleviate the data scarcity limitation, we design Sequence Decomposition Adaptation Network (SeDAN) which is a novel transfer architecture to improve forecasting performance on the target domain by aligning transferable knowledge from cross-domain datasets. Rethinking the transferability of features in time series data, we propose Implicit Contrastive Decomposition to decompose the original features into components including seasonal and trend features, which are easier to transfer. Then we design the corresponding adaptation methods for decomposed features in different domains. Specifically, for seasonal features, we perform joint distribution adaptation and for trend features, we design an Optimal Local Adaptation. We conduct extensive experiments on five benchmark datasets for multivariate time series forecasting. The results demonstrate the effectiveness of our SeDAN. It can provide more efficient and stable knowledge transfer.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
