Multicomponent Spatial-Temporal Graph Attention Convolution Networks for Traffic Prediction with Spatially Sparse Data

Predicting traffic data on traffic networks is essential to transportation management. It is a challenging task due to the complicated spatial-temporal dependency. The latest studies mainly focus on capturing temporal and spatial dependencies with spatially dense traffic data. However, when traffic data become spatially sparse, existing methods cannot capture sufficient spatial correlation information and thus fail to learn the temporal periodicity sufficiently. To address these issues, we propose a novel deep learning framework, Multi-component Spatial-Temporal Graph Attention Convolutional Networks (MSTGACN), for traffic prediction, and we successfully apply it to predicting traffic flow and speed with spatially sparse data. MSTGACN mainly consists of three independent components to model three types of periodic information. Each component in MSTGACN combines dilated causal convolution, graph convolution layer, and the weight-shared graph attention layer. Experimental results on three real-world traffic datasets, METR-LA, PeMS-BAY, and PeMSD7-sparse, demonstrate the superior performance of our method in the case of spatially sparse data.

Transient hexagonal structures in sheared emulsions of isotropic inclusions on smectic bubbles in microgravity conditions

We describe the collective behavior of isotropic droplets dispersed over a spherical smectic bubble, observed under microgravity conditions on the International Space Station (ISS). We find that droplets can form two-dimensional hexagonal structures changing with time. Our analysis indicates the possibility of spatial and temporal periodicity of such structures of droplets. Quantitative analysis of the hexagonal structure including the first three coordination circles was performed. A peculiar periodic-in-time ordering of the droplets, related to one-dimensional motion of droplets with non-uniform velocity, was found.

Real-time CMR Fourier Analysis for Temporospatial Characterization of Ventricular Wall Motion in Health and Disease

Background Cardiovascular magnetic resonance (CMR) has been largely dependent on retrospective cine for image acquisition. Real-time imaging, although inferior in image quality to retrospective cine, is advantageous in examining temporospatial behaviors of cardiac motion over a series of sequential cardiac cycles. The presented work is a proof-of-concept of assessing cardiac function quantitatively with novel temporospatial indices in real-time CMR. Methods Fourier analysis was introduced for temporospatial characterization of real-time CMR signals arising from ventricular wall motion. Two quantitative indices, temporal periodicity and spatial coherence, were provided for function assessment in the left ventricle (LV) and right ventricle (RV). We prospectively investigated these temporospatial indices in a CMR study with healthy volunteers and heart failure (HF) patients. Results Real-time images were collected and analyzed in 12 healthy volunteers during exercise and at rest, and also in 12 HF patients at rest. The statistics indicated that the healthy volunteers presented an increase of temporal periodicity due to ventricular response to exercise (resting-state 0.24 ± 0.037 vs. exercising-state 0.31 ± 0.040 in LV; resting-state 0.18 ± 0.030 vs. exercising-state 0.25 ± 0.038 in RV; P < 0.001 for both). The HF patients gave lower temporal periodicity (0.14 ± 0.021 for LV; 0.10 ± 0.014 for RV; P < 0.001 for both) than that in the healthy volunteers. The spatial coherence of LV and RV wall motion was also lower in the HF patients than that in the healthy volunteers (0.38 ± 0.040 vs 0.52 ± 0.039 for LV; 0.35 ± 0.035 vs. 0.50 ± 0.036 for RV; P < 0.001 for both). Both temporal periodicity and spatial coherence were found to be correlated to end-systolic volume (ESV) and ejection-fraction (EF) (R > 0.6, P < 0.001). However, the HF patients and healthy volunteers were well differentiated in the scatter plots of spatial coherence and temporal periodicity while they were mixed in those of ESV and EF. Conclusions Real-time CMR Fourier analysis enables a new approach to quantitative assessment of cardiac function with temporal periodicity and spatial coherence. The temporospatial characterization of real-time CMR images has the potential for intricate analysis of ventricular wall motion beyond conventional methods.

Location Prediction over Sparse User Mobility Traces Using RNNs: Flashback in Hidden States!

Location prediction is a key problem in human mobility modeling, which predicts a user's next location based on historical user mobility traces. As a sequential prediction problem by nature, it has been recently studied using Recurrent Neural Networks (RNNs). Due to the sparsity of user mobility traces, existing techniques strive to improve RNNs by considering spatiotemporal contexts. The most adopted scheme is to incorporate spatiotemporal factors into the recurrent hidden state passing process of RNNs using context-parameterized transition matrices or gates. However, such a scheme oversimplifies the temporal periodicity and spatial regularity of user mobility, and thus cannot fully benefit from rich historical spatiotemporal contexts encoded in user mobility traces. Against this background, we propose Flashback, a general RNN architecture designed for modeling sparse user mobility traces by doing flashbacks on hidden states in RNNs. Specifically, Flashback explicitly uses spatiotemporal contexts to search past hidden states with high predictive power (i.e., historical hidden states sharing similar contexts as the current one) for location prediction, which can then directly benefit from rich spatiotemporal contexts. Our extensive evaluation compares Flashback against a sizable collection of state-of-the-art techniques on two real-world LBSN datasets. Results show that Flashback consistently and significantly outperforms state-of-the-art RNNs involving spatiotemporal factors by 15.9% to 27.6% in the next location prediction task.

Towards Efficient Mining of Periodic High-Utility Itemsets in Large Databases

High Utility Item sets mining has attracted many researchers in recent years. But HUI mining methods involves a exponential mining space and returns a very large number of high-utility itemsets. . Temporal periodicity of itemset is considered recently as an important interesting criteria for mining high-utility itemsets in many applications. Periodic High Utility item sets mining methods has a limitation that it does not consider frequency and not suitable for large databases. To address this problem, we have proposed two efficient algorithms named FPHUI( mining periodic frequent HUIs), MFPHM(efficient mining periodic frequent HUIs) for mining periodic frequent high-utility itemsets. The first algorithm FPHUI miner generates all periodic frequent itemsets. Mining periodic frequent high-utility itemsets leads to more computational cost in very large databases. We further developed another algorithm called MFPHM to overcome this limitation. The performance of the frequent FPHUI miner is evaluated by conducting experiments on various real datasets. Experimental results show that proposed algorithms is efficient and effective.

On the approximation of spatial structures of global tidal magnetic field models

The extraction of the magnetic signal induced by the oceanic M2 tide is typically based solely on the temporal periodicity of the signal. Here, we propose a system of tailored trial functions that additionally takes the spatial constraint into account that the sources of the signal are localized within the oceans. This construction requires knowledge of the underlying conductivity model but not of the inducing tidal current velocity. Approximations of existing tidal magnetic field models with these trial functions and comparisons with approximations based on other localized and nonlocalized trial functions are illustrated.

On the Approximation of Spatial Structures of Global Tidal Magnetic Field Models

The extraction of the magnetic signal induced by the oceanic M2 tide is typically based solely on the temporal periodicity of the signal. Here, we propose a system of tailored trial functions that additionally takes the spatial constraint into account that the sources of the signal are localized within the oceans. This construction requires knowledge of the underlying conductivity model but not of the inducing tidal current velocity. Approximations of existing tidal magnetic field models with these trial functions and comparisons with approximations based on other localized and non-localized trial functions are illustrated.