Modeling complex spatiotemporal dependencies in correlated traffic series is essential for prediction. While recent works have shown improved prediction performance by using neural networks to extract correlations, their effectiveness depends on the quality of graph structures used represent spatial topology network. In this work, we propose a novel approach that embeds time-varying dynamic Bayesian network capture fine data. We then use convolutional generate forecasts. To enable our method...

Reinforcement learning (RL) can automatically learn a better policy through trial-and-error paradigm and has been adopted to revolutionize optimize traditional traffic signal control systems that are usually based on handcrafted methods. However, most existing RL-based models either single scenario or multiple independent scenarios, where each separate simulation environment with predefined road network topology settings. These implement training testing in the same scenario, thus being...

Low-rank tensor decomposition and completion have attracted significant interest from academia given the ubiquity of data. However, low-rank structure is a global property, which will not be fulfilled when data presents complex weak dependencies specific graph structures. One particular application that motivates this study spatiotemporal analysis. As shown in preliminary study, weakly can worsen performance. In paper, we propose novel CANDECOMP / PARAFAC (CP) framework by introducing...

Road network and trajectory representation learning are essential for traffic systems since the learned can be directly used in various downstream tasks (e.g., speed inference, travel time estimation). However, most existing methods only contrast within same scale, i.e., treating road separately, which ignores valuable inter-relations. In this paper, we aim to propose a unified framework that jointly learns representations end-to-end. We design domain-specific augmentations road-road...

The effectiveness of traffic light control has been significantly improved by current reinforcement learning-based approaches via better cooperation among multiple lights. However, a persisting issue remains: how to obtain multi-agent signal algorithm with remarkable transferability across diverse cities? In this paper, we propose Transformer on (TonT) model for cross-city meta control, named as X-Light: We input the full Markov Decision Process trajectories, and Lower aggregates states,...

Background: Pancreatic ductal adenocarcinoma (PDAC) features a complex tumor microenvironment (TME) that significantly influences patient outcomes. Understanding the TME's cellular composition and interactions is crucial for identifying therapeutic targets to improve treatment. Methods We performed an integrative analysis combining 88 single-cell RNA sequencing (scRNA-seq) samples with 187,520 cells, 20 Visium spatial transcriptomics (ST-seq) 67,933 spots, 1383 bulk RNA-seq 2 Xenium...

Sensors are commonly deployed to perceive the environment. However, due high cost, sensors usually sparsely deployed. Kriging is tailored task infer unobserved nodes (without sensors) using observed (with sensors). The essence of kriging transferability. Recently, several inductive spatio-temporal methods have been proposed based on graph neural networks, being trained a built top via pretext tasks such as masking out and reconstructing them. in training inevitably much sparser than...

For young children, family meals are an enjoyable and developmentally useful part of daily life. Although prior work has shown that ubiquitous computing solutions can enhance children's eating habits mealtime experiences in valuable ways, other demonstrates many families hesitant to use technology this context. This paper examines adoption barriers for understand with more nuance what parents value resist space. Using mixed methods, we first observed dinnertime then surveyed 122 children...

Effective multi-intersection collaboration is pivotal for reinforcement-learning-based traffic signal control to alleviate congestion. Existing work mainly chooses neighboring intersections as collaborators. However, quite a lot of congestion, even some wide-range caused by non-neighbors failing collaborate. To address these issues, we propose separate the collaborator selection second policy be learned, concurrently being updated with original signal-controlling policy. Specifically, in...

Spatiotemporal data are very common in many applications, such as manufacturing systems and transportation systems. Given the intrinsic complex spatial temporal correlations of data, short-term long-term prediction for spatiotemporal is often challenging. Most traditional statistical models fail to preserve innate features alongside their correlations. In this paper, we focus on a tensor-based method propose several practical techniques improve both accuracy. For prediction, "tensor...

Anomaly detection is an essential task for quality management in smart manufacturing. An accurate data-driven method usually needs enough data and labels. However, practice, there commonly exist newly set-up processes manufacturing, they only have quite limited available analysis. Borrowing the name from recommender system, we call this process a cold-start process. The sparsity of anomaly, deviation profile, noise aggravate difficulty. Transfer learning could help to detect anomalies by...

Abstract Individual passenger travel patterns have significant value in understanding passenger’s behavior, such as learning the hidden clusters of locations, time, and passengers. The learned further enable commercially beneficial actions customized services, promotions, data-driven urban-use planning, peak hour discovery, so on. However, individualized modeling is very challenging for following reasons: 1) individual data are multi-dimensional spatiotemporal big data, including at least...

Abstract Artificial intelligence (AI) systems have been widely applied to various contexts, including high-stake decision processes in healthcare, banking, and judicial systems. Some developed AI models fail offer a fair output for specific minority groups, sparking comprehensive discussions about fairness. We argue that the development of is marked by central paradox: less participation one stakeholder has within system’s life cycle, more influence they over way system will function. This...

Large Language Models (LLMs) have emerged as a powerful tool in advancing the Text-to-SQL task, significantly outperforming traditional methods. Nevertheless, nascent research field, there is still no consensus on optimal prompt templates and design frameworks. Additionally, existing benchmarks inadequately explore performance of LLMs across various sub-tasks process, which hinders assessment LLMs' cognitive capabilities optimization LLM-based solutions. To address aforementioned issues, we...

Traffic congestion is becoming an increasingly prominent problem, and intelligent traffic signal control methods can effectively alleviate it. Recently, there has been a growing trend of applying reinforcement learning to for adaptive scheduling. However, most existing focus on improving performance while neglecting the issue scheduling fairness, resulting in long waiting time some vehicles. Some works attempt address fairness issues but often sacrifice transport performance. Furthermore,...

This paper proposes to learn Multi-task, Multi-modal Direct Acyclic Graphs (MM-DAGs), which are commonly observed in complex systems, e.g., traffic, manufacturing, and weather whose variables multi-modal with scalars, vectors, functions. takes the traffic congestion analysis as a concrete case, where intersection is usually regarded DAG. In road network of multiple intersections, different intersections can only have someoverlapping distinct observed. For example, signalized has...

Traffic prediction, a critical component for intelligent transportation systems, endeavors to foresee future traffic at specific locations using historical data. Although existing prediction models often emphasize developing complex neural network structures, their accuracy has not seen improvements accordingly. Recently, Large Language Models (LLMs) have shown outstanding capabilities in time series analysis. Differing from models, LLMs progress mainly through parameter expansion and...

Recent advancements in Text-to-SQL (Text2SQL) emphasize stimulating the large language models (LLM) on in-context learning, achieving significant results. Nevertheless, they face challenges when dealing with verbose database information and complex user intentions. This paper presents a two-stage framework to enhance performance of current LLM-based natural SQL systems. We first introduce novel prompt representation, called reference-enhanced which includes schema randomly sampled cell...

Existing multi-agent deep reinforcement learning (MADRL) methods for multi-UAV navigation face challenges in generalization, particularly when applied to unseen complex environments. To address these limitations, we propose a Dual-Transformer Encoder-Based Proximal Policy Optimization (DTPPO) method. DTPPO enhances collaboration through Spatial Transformer, which models inter-agent dynamics, and Temporal captures temporal dependencies improve generalization across diverse This architecture...

Low-rank tensor decomposition and completion have attracted significant interest from academia given the ubiquity of data. However, low-rank structure is a global property, which will not be fulfilled when data presents complex weak dependencies specific graph structures. One particular application that motivates this study spatiotemporal analysis. As shown in preliminary study, weakly can worsen performance. In paper, we propose novel CANDECOMP / PARAFAC (CP) framework by introducing...

Individualized passenger travel pattern is of significant research value since the abundant information from individual trajectory data could help discover useful insights about multi-clustering origin, destination, time, etc., and cluster. However, this task rather challenging, given high-dimensional, with complex spatiotemporal structure, exhibits sparse patterns. Moreover, patterns are also affected by external information, such as distances functions locations points interest, which...

Correlated time series analysis plays an important role in many real-world industries. Learning efficient representation of this large-scale data for further downstream tasks is necessary but challenging. In paper, we propose a time-step-level learning framework individual instances via bootstrapped spatiotemporal prediction. We evaluated the effectiveness and flexibility our on correlated forecasting cold-start transferring model to new with limited data. A linear regression trained top...