The recovery of coarsely sampled trajectories considering the road network topology characteristics is a crucial task for many downstream applications in intelligent transportation systems. Existing approaches in this domain primarily focus on extracting spatio-temporal correlations for the observed trajectory points but neglect the critical role of road network topology characteristics in making the recovery results more accurate and realistic. In addition, too many road segments in cities undermine the model inference performance. To address these challenges, we propose a novel Map-informed Adaptive Spatio-Temporal Autoencoder, which follows an encoder-decoder architecture for trajectory recovery. Specifically, we utilize a pre-trained attributed network embedding module to incorporate the road segment characteristics into the input data to make it easier for the model to extract the spatio-temporal dependencies from coarse trajectories. Furthermore, we construct a novel adaptive mask inference module that contains a distance-based mask matrix and a learnable adaptive mask matrix to assist the model in making segment inferences by weighting each candidate segment adaptively in the recovery process. To evaluate the performance of the proposed model, we conduct a series of comprehensive case studies on two representative real-world trajectory datasets. The experimental results demonstrate that the proposed model consistently outperforms state-of-the-art approaches.

Exploiting spatial–temporal correlations has long been regarded as the cornerstone of traffic state prediction. Among existing techniques, temporal graph neural networks (TGNNs) have recently emerged as a prominent solution for modeling complex spatial–temporal traffic data correlations. Existing studies on TGNNs mainly focus on developing new building blocks to embed hidden correlations into a unified latent representation, which is mapped to predictions of distinct horizons. However, mapping the same latent features to distinct scalar predictions makes the gradient computation challenging for updating model parameters in the relevant directions. Besides, TGNNs are biased towards the shared temporal patterns while neglecting the complex dependencies within each data series, which can be captured to enrich latent features. To handle these problems jointly, we propose a novel feature projection scheme for the traffic prediction framework of TGNNs. The proposed projection scheme is based on spatial convolutions that first generate horizon-specific feature maps and then transform them into scalar predictions of the corresponding horizons. These horizon-specific feature maps establish interactions between the unified latent representation and the corresponding output values to bring the predictions closer to the true values. Besides, the proposed scheme also serves as a pattern modeling phase that enhances the expressivity of TGNNs by enriching latent features with data source-wise patterns of distinct time steps. Comprehensive experiments on two real-world traffic datasets demonstrate that the proposed scheme enhances the predictive performance and reduces the model parameters of TGNNs.

Deep neural networks (DNNs) have emerged as a dominant approach for developing traffic forecasting models. These models are typically trained to minimize error on averaged test cases and produce a single-point prediction, such as a scalar value for traffic speed or travel time. However, single-point predictions fail to account for prediction uncertainty that is critical for many transportation management scenarios, such as determining the best-or worst-case arrival time. We present, a generic framework to enhance the capability of an arbitrary DNN model for uncertainty modeling. requires little human involvement and does not change the base DNN architecture during deployment. Instead, it automatically learns a standard quantile function during the DNN model training to produce a prediction interval for the single-point prediction. The prediction interval defines a range where the true value of the traffic prediction is likely to fall. Furthermore, develops an adaptive scheme that dynamically adjusts the prediction interval based on the location and prediction window of the test input. We evaluated by applying it to five representative DNN models for traffic forecasting across seven public datasets. We then compared against six uncertainty quantification methods. Compared to the baseline uncertainty modeling techniques, with base DNN architectures delivers consistently better and more robust performance than the existing ones on the reported datasets.

Personalized recipe recommender systems help users mine certain dishes they want to find and even really desire, which play a significant role in matching dishes, balancing nutrients, and preventing non-communicable diseases. Generally, customer’s preferences or needs vary from person to person, and people are often reluctant to accept recommended food without reasonable explanation, especially when their demands are not explicitly addressed. In this paper, we are devoted to providing recipe suggestions accompanied by rational interpretations generated from images or videos. First, we construct a recipe knowledge graph (RcpKG) through the use of multi-modality and hierarchical thought, which focuses on the underlying demands of users and the consideration of multiple fine-grained factors. On this basis, a novel multi-modal recipe recommendation method via the knowledge graph (RcpMKR) is proposed, which represents nodes in multiple aspects and performs multi-relational graph structure extraction of the RcpKG. It not only takes into account local associations within the graph but also global information, and incorporates user concerns at different levels. Then, we adopt BERT-based multi-modal models and generative adversarial networks to generate interpretations. Additionally, dynamic convolution and random synthetic attention are utilized in our work to discriminate among features. Experimental results show that the proposed method and BERT-based fusion models improve recipe recommendation performance and explanation generation. Specifically, the precision of the RcpMKR method through RcpKG, user concerns and graph convolutional network improves by 7.82%, and the viExpCBTBERT method via 2D&3D convolutional neural networks for developing text interpretations enhances the F1-score by 10% compared with the baseline.