基于半监督宽度学习系统的气温空间插值

近年来,基于人工神经网络的气象空间插值研究因其较强的非线性映射能力受到越来越多的关注。然而,由于网络参数众多,这些方法的训练过程都非常耗时。考虑到以上原因,基于宽度学习系统提出了一个高效的气温空间插值模型,首先使用稀疏编码对输入数据进行特征提取,然后将映射后的向量引入宽度学习系统中得到一个半监督宽度学习模型。相关实验建立在山西省的气象数据集上,实验结果表明,该方法比基于深度学习的方法和传统空间插值方法具有更好的预测性能和时间损耗。

Air Temperature Spatial Interpolation Based On Semi-Supervised Broad Learning System

Recently, artificial neural networks have drawn increasing attention in meteorological spatial interpolation, due to their strong nonlinear mapping capability. However, their methods suffer from a time-consuming training process because of many network parameters. Considering the above reasons, an efficient air temperature spatial interpolation model based on the broad learning system is proposed, firstly, sparse coding is used to extract features of input data, and then the mapped vector is incorporated into broad learning system to obtain a semi-supervised broad learning version. Relevant experiments are established on meteorological data sets of Shanxi Province , experimental results verified the proposed algorithm possesses a better performance and time loss compared to deep learning-based methods and the conventional spatial interpolation methods.