This study aims to explore the possibility of using a multilayer artificial neural network for the classification between children with autism spectrum disorder (ASD) and typically developing (TD) children based on short-time spontaneous hemodynamic fluctuations. Spontaneous hemodynamic fluctuations were collected by a functional near-infrared spectroscopy setup from bilateral inferior frontal gyrus and temporal cortex in 25 children with ASD and 22 TD children. To perform feature extraction and classification, a multilayer neural network called CGRNN was used which combined a convolution neural network (CNN) and a gate recurrent unit (GRU), since CGRNN has a strong ability in finding characteristic features and acquiring intrinsic relationship in time series. For the training and predicting, short-time (7 s) time-series raw functional near-infrared spectroscopy (fNIRS) signals were used as the input of the network. To avoid the over-fitting problem and effectively extract useful differentiation features from a sample with a very limited size (e.g., 25 ASDs and 22 TDs), a sliding window approach was utilized in which the initially recorded long-time (e.g., 480 s) time-series was divided into many partially overlapped short-time (7 s) sequences. By using this combined deep-learning network, a high accurate classification between ASD and TD could be achieved even with a single optical channel, e.g., 92.2% accuracy, 85.0% sensitivity, and 99.4% specificity. This result implies that the multilayer neural network CGRNN can identify characteristic features associated with ASD even in a short-time spontaneous hemodynamic fluctuation from a single optical channel, and second, the CGRNN can provide highly accurate prediction in ASD.

Sea surface temperature (SST) research is the basis of our understanding of local and global climate characteristics. Significant effort has been focused on the acquisition of accurate prediction results. Although a large amount of SST data is available due to the development of storage media, there are limited studies on the prediction of future SST values in a spatiotemporal manner. Most previous studies ignore the regional distribution information in natural geographical location, which leads to the loss of regional information and hinders further regional analysis. In this work, we have addressed this issue by presenting Regional Convolution long short-term memory (RC-LSTM) model with spatio-temporal information processing capability. To the best of our knowledge, this is the first time that SST time series has been used for region prediction based on deep learning. Daily SST from January 2000 to December 2012 from China’s offshore waters is adopted to conduct experiments. The local correlation and global coherence of each time series are extracted by the multi-layer neural network in the model. The structure of the RC-LSTM model includes a convolutional layer with region information extraction capabilities, and a long short-term memory (LSTM) layer. The RC-LSTM model performs well over both the South China Sea and the East China Sea, as determined by the root mean squared error (RMSE) and the accuracy (Acc). The experimental results show that our model is more accurate than traditional prediction models and has higher generalization ability, and is therefore more practical.

The study of sea surface temperature (SST) in coastal water is of great significance for navigation, aquaculture and military. Numerous studies have been conducted to predict this parameter in recent years. The fluctuation of SST is periodic, and it shows different changing patterns over different timescales. At present, most investigations on SST ignore the influence of multiscale features on the prediction, which may limit the accuracy of the final prediction. To fully exploit the features of SST data, we propose a multi-long short-term memory convolution neural network (M-LCNN) prediction model. In this model, we use the wavelet transform to decompose and reconstruct the time series, we then predict the variation of SST sequences at multiple scales, and finally complete the prediction process. We conduct experiments in the Yellow Sea and the Bohai Sea in China, and the results indicate that our method is significantly better than traditional approaches.