Temporal Knowledge Graphs (TKGs) model dynamic events, and understanding temporal evolution is crucial for effective reasoning. While existing methods leverage Graph Neural Networks (GNNs) to model structural dependencies, they often rely on Recurrent Neural Networks (RNNs) to process sequence of graph structures. They struggle to (1) incorporate contextual information, (2) explicitly model long-term effects, and (3) handle in-equidistant data. To address these challenges, we propose Context-Aware Hawkes Graph Network (CoHN), a novel TKG reasoning approach based on the Hawkes process. CoHN features a tailored conditional intensity function that models TKG event occurrences. It is characterized by two additive terms: base intensity and historical influence, representing spontaneous tendency and influence of past events at in-equidistant time intervals. Firstly, we design a Contextual Encoder (CE) to encode contextual information for all entities and compute the base intensity. We then present an attention-based Evolutionary Encoder that captures local structural information and explicitly models long-term dependencies across the TKG. A self-exciting fusion module further aggregates historical evolutionary dependencies at all timestamps to quantify the final historical influence. Extensive experiments on common benchmarks demonstrate the superiority, robustness, and efficiency of our method.

Human image generation is a very challenging task since it is affected by many factors. Many human image generation methods focus on generating human images conditioned on a given pose, while the generated backgrounds are often blurred. In this paper, we propose a novel Partition-Controlled GAN to generate human images according to target pose and background. Firstly, human poses in the given images are extracted, and foreground/background are partitioned for further use. Secondly, we extract and fuse appearance features, pose features and background features to generate the desired images. Experiments on Market-1501 and DeepFashion datasets show that our model not only generates realistic human images but also produce the human pose and background as we want. Extensive experiments on COCO and LIP datasets indicate the potential of our method.

Text detection in natural scene images is an important and popular task in the computer vision community. Due to slanted characters and blurred images in natural environments, it is a challenging task under active research. In this paper, we propose a Focal Text Detection Network (FTDN), which could be trained well without abundant data. FTDN is able to segment text region and simultaneously regress text box at pixel-level. Specifically, combined with focal loss, our method can balance positive/negative and easy/hard samples to achieve better performance. Compared with previous methods, FTDN achieves better performance in terms of text detection accuracy in natural scene. It outperforms the state-of-the-art methods on the standard ICDAR 2015 dataset with 80.9% F-measure.

The authors present a numerically efficient mathematical approach, the numerical integral method, for diffusion modeling. This method extends the boundary integral methods to solve linear and nonlinear diffusion problems with physically based diffusivity models to account for concentration-dependent effects, oxidizing ambient effects, electric field effects, etc. The algorithm for solving nonlinear diffusion problems and its implementation in a computer program are described. Results from example problems are shown. This new numerical method provides the potential for achieving a diffusion simulation which is computationally efficient, numerically stable, and sufficiently accurate in an engineering environment. Such an efficient and yet accurate diffusion simulation is especially important in statistical simulation of VLSI processes.>