Search

Precipitation nowcasting, which involves the short-term, high-resolution prediction of rainfall, plays a crucial role in various real-world applications. In recent years, researchers have increasingly utilized deep learning-based methods in precipitation nowcasting. The exponential growth of spatiotemporal observation data has heightened interest in recent advancements such as denoising diffusion models, which offer appealing prospects due to their inherent probabilistic nature that aligns well with the complexities of weather forecasting. Successful application of diffusion models in rainfall prediction tasks requires relevant conditions and effective utilization to direct the forecasting process of the diffusion model. In this paper, we propose a probabilistic spatiotemporal model for precipitation nowcasting, named LLMDiff. The architecture of LLMDiff includes two networks: a conditional encoder–decoder network and a denoising network. The conditional network provides conditional information to guide the denoising network for high-quality predictions related to real-world earth systems. Additionally, we utilize a frozen transformer block from pre-trained large language models (LLMs) in the denoising network as a universal visual encoder layer, which enables the accurate estimation of motion trend by considering long-term temporal context information and capturing temporal dependencies within the frame sequence. Our experimental results demonstrate that LLMDiff outperforms state-of-the-art models on the SEVIR dataset. [ABSTRACT FROM AUTHOR]

Short-term precipitation forecasting is essential for agriculture, transportation, urban management, and tourism. The radar echo extrapolation method is widely used in precipitation forecasting. To address issues like forecast degradation, insufficient capture of spatiotemporal dependencies, and low accuracy in radar echo extrapolation, we propose a new model: MS-DD3D-RSTN. This model employs spatiotemporal convolutional blocks (STCBs) as spatiotemporal feature extractors and uses the spatial-temporal loss (STLoss) function to learn intra-frame and inter-frame changes for end-to-end training, thereby capturing the spatiotemporal dependencies in radar echo signals. Experiments on the Sichuan dataset and the HKO-7 dataset show that the proposed model outperforms advanced models in terms of CSI and POD evaluation metrics. For 2 h forecasts with 20 dBZ and 30 dBZ reflectivity thresholds, the CSI metrics reached 0.538, 0.386, 0.485, and 0.198, respectively, representing the best levels among existing methods. The experiments demonstrate that the MS-DD3D-RSTN model enhances the ability to capture spatiotemporal dependencies, mitigates forecast degradation, and further improves radar echo prediction performance. [ABSTRACT FROM AUTHOR]

Vascular diseases, such as venous insufficiency and coronary artery diseases, have been threatening the health of people. Efficient treatment with proper postoperative care is required to relieve the pain of the patients. Traditionally, venous insufficiency is treated with ligation and stripping, an open surgery whose complication rate cannot be ignored. Coronary artery disease is often treated with balloon angioplasty during which undilatable lesions may be encountered, limiting the efficacy of this approach. With advances in laser photonics and percutaneous coronary intervention procedure, laser ablation is emerging as an alternative and adjunctive therapy for these diseases. Endovenous laser ablation has the advantages of high success rate, low complication risk, and fast postoperative recovery. Laser ablation in arteries can handle uncrossable or undilatable lesions with a low incidence of serious complications. In this review, previously published research concerning vascular diseases and their therapies are analyzed in order to provide a clear explanation of the mechanisms and merits of laser ablation. For endovenous laser ablation, the main mechanisms are steam bubbles, heat conduction, and heat pipe, and three main influencing factors are wavelength, fiber types, and laser energy density. For excimer laser coronary atherectomy, the main mechanisms are photochemical, photothermal, and photomechanical effects, and three main influencing factors are catheter, medium, and laser parameters. [ABSTRACT FROM AUTHOR]

A Theory-Driven Deep Learning Method for Voice Chat–Based Customer Response Prediction In this study, we target the task of voice chat–based customer response prediction in an emerging online interaction–based commercial mode, the invite-online-and-experience-in-store mode. Prior research shows that satisfaction, which can be revealed by the discrepancy between prior expectation and actual experience, is a key factor to disentangle customers' purchase intention, whereas black box deep learning methods empirically promise us with advantageous capabilities in dealing with complex voice data, for example, text and audio information incorporated in voice chat. To this end, we propose a theory-driven deep learning method that enables us to (1) learn customers' personalized product preferences and dynamic satisfaction in the absence of their profile information, (2) model customers' actual experiences based on multiview voice chat information in an interlaced way, and (3) enhance the customer response prediction performance of a black box deep learning model with theory-driven dynamic satisfaction. Empirical evaluation results demonstrate the advantageous prediction performance of our proposed method over state-of-the-art deep learning alternatives. Investigation of cumulative satisfaction reveals the collaborative predictive roles of theory-driven dynamic satisfaction and deep representation features for customer response prediction. Explanatory analysis further renders insights into customers' personalized preferences and dynamic satisfaction for key product attributes. As artificial intelligence and digitalization technologies are flourishing real-time, online interaction–based commercial modes, exploiting customers' purchase intention implied in online interaction processes may foster huge business opportunities. In this study, we target the task of voice chat–based customer response prediction in an emerging online interaction–based commercial mode, the invite-online-and-experience-in-store mode. Prior research shows that satisfaction, which can be revealed by the discrepancy between prior expectation and actual experience, is a key factor to disentangle customers' purchase intention, whereas black-box deep learning methods empirically promise us advantageous capabilities in dealing with complex voice data, for example, text and audio information incorporated in voice chat. To this end, we propose a theory-driven deep learning method that enables us to (1) learn customers' personalized product preferences and dynamic satisfaction in the absence of their profile information, (2) model customers' actual experiences based on multiview voice chat information in an interlaced way, and (3) enhance the customer response prediction performance of a black-box deep learning model with theory-driven dynamic satisfaction. Empirical evaluation results demonstrate the advantageous prediction performance of our proposed method over state-of-the-art deep learning alternatives. Investigation of cumulative satisfaction reveals the collaborative predictive roles of theory-driven dynamic satisfaction and deep representation features for customer response prediction. Explanatory analysis further renders insights into customers' personalized preferences and dynamic satisfaction for key product attributes. History: Yong Tan, Senior Editor; Jingjing Zhang, Associate Editor. Funding: This work was supported in part by the National Natural Science Foundation of China [Grants 71971067 and 72271059] and the China Postdoctoral Science Foundation [Grant 2022M722394]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/isre.2022.1196. [ABSTRACT FROM AUTHOR]