Finding top-k items in data streams is a fundamental problem mining. Existing algorithms that can achieve unbiased estimation suffer from poor accuracy. In this paper, we propose new sketch, WavingSketch, which much more accurate than existing algorithms. WavingSketch generic, and show how it be applied to four applications: finding frequent items, heavy changes, persistent Super-Spreaders. We theoretically prove provide estimation, then give an error bound of our algorithm. Our experimental...

At present, internet celebrity marketing has become a driving force for the growth of mobile e-commerce; however, it also more apparent that credibility and authenticity is directly correlated to success model. Therefore, in order entice consumers into purchasing products, cooperations celebrities must be deemed trustworthy. In addition, there are several factors influence trust between consumers. To highlight these factors, this paper constructed an model from perspective takes features,...

Transform-domain image denoising methods assume that the original signal can be sparsely represented in transform domain, but none of orthogonal transforms achieve sparse representation for all images. Proposed is a hybrid Fourier-wavelet method to overcome this shortcoming. Experimental results show proposed algorithm improves performance efficiently.

High-accuracy real-time data stream estimations are critical for various applications, and sliding-window-based techniques have attracted wide attention. However, existing solutions struggle to achieve high accuracy, generality, low memory usage simultaneously. To overcome these limitations, we present MicroscopeSketch, a high-accuracy sketch framework. Our key technique, called adaptive zooming, dynamically adjusts the granularity of counters maximize accuracy while minimizing usage. By...

In the study of spatiotemporal data visualization, compression and morphing density maps are challenging tasks. Many existing methods require adjustment multiple parameters rich experience, but still cannot get accuracy or smooth results. this paper, we propose a GAN-based method (LatentMap) to explore latent space maps, which is an end-to-end method. First, find that small codes can be used as results compression, greatly save transmission time in front-end system. We collect make dataset....

The Aesthetics Assessment of Children's Paintings (AACP) is an important branch the image aesthetics assessment (IAA), playing a significant role in children's education. This task presents unique challenges, such as limited available data and requirement for evaluation metrics from multiple perspectives. However, previous approaches have relied on training large datasets subsequently providing score to image, which not applicable AACP. To solve this problem, we construct dataset paintings...

The Aesthetics Assessment of Children's Paintings (AACP) is an important branch the image aesthetics assessment (IAA), playing a significant role in children's education. This task presents unique challenges, such as limited available data and requirement for evaluation metrics from multiple perspectives. However, previous approaches have relied on training large datasets subsequently providing score to image, which not applicable AACP. To solve this problem, we construct dataset paintings...

Abstract Visual querying of spatiotemporal data has become a dominant mode in the field visual analytics. Previous studies have utilized well‐designed structures to speed up data. However, reducing storage overhead while improving efficiency distribution remains significant challenge. We propose flow‐based neural representation method for efficient querying. First, we transform into density maps through kernel estimation. Then, leverage data‐driven modeling capabilities network achieve...

By converting low-frame-rate, low-resolution videos into high-frame-rate, high-resolution ones, space-time video super-resolution techniques can enhance visual experiences and facilitate more efficient information dissemination. We propose a convolutional neural network (CNN) for super-resolution, namely GIRNet. To generate highly accurate features thus improve performance, the proposed integrates feature-level temporal interpolation module with deformable convolutions global...

We propose an end-to-end attribute compression method for dense point clouds. The proposed combines a frequency sampling module, adaptive scale feature extraction module with geometry assistance, and global hyperprior entropy model. uses Hamming window the Fast Fourier Transform to extract high-frequency components of cloud. difference between original cloud sampled is divided into multiple sub-point These clouds are then partitioned using octree, providing structured input extraction....

In network measurement, sliding window measurement has the advantage of providing recent and timely results. Recently, sketches have become most popular method conducting flow-level measurements due to their favorable trade-off between small memory overhead high accuracy. However, it remains a challenge that no current are able support unbiased estimation toward flow size which can improve performance tasks including diagnoses, delay heavy hitter detection. this paper, we propose first work...

Multi-set membership queries are fundamental operations in data science. In this paper, we propose a new structure for multi-set queries, named coloring embedder, which is fast, accurate, and memory efficient. The idea of embedder to first map elements high-dimensional space, nearly eliminates hashing collisions, then use dimensional reduction representation, similar graph, save memory. Theoretical proofs experimental results show that the effective solving problem queries. We also find web...

Deep learning models are increasingly deployed to edge devices for real-time applications. To ensure stable service quality across diverse environments, it is highly desirable generate tailored model architectures different conditions. However, conventional pre-deployment generation approaches not satisfactory due the difficulty of handling diversity environments and demand information. In this paper, we propose adapt architecture after deployment in target environment, where can be...

The evaluation of a company's value can serve as guide for investors to assess the company and make informed investment decisions. However, conventional valuation techniques are not applicable Initial Public Offering (IPO) companies in China, mainly due absence historical market performance. In contrast, finance statement provides periodic overview operational production activities, which is linked its Traditional methods often rely on selection limited number financial indicators from...

Image style transfer is a challenging task in computational vision. Existing algorithms the color and texture of images by controlling neural network's feature layers. However, they fail to control strength textures different regions content image. To address this issue, we propose training method that uses loss function constrain intensity regions. This guides features based on gradient relationship between images. Additionally, introduce novel fusion linearly transforms resemble while...

Edge computing is being widely used for video analytics. To alleviate the inherent tension between accuracy and cost, various analytics pipelines have been proposed to optimize usage of GPU on edge nodes. Nonetheless, we find that compute resources provisioned nodes are commonly under-utilized due content variations, subsampling filtering at different places a pipeline. As opposed model pipeline optimization, in this work, study problem opportunistic data enhancement using non-deterministic...

An error-controlled iterative algorithm (ECIA) with extra two thresholds to reduce decision errors is proposed for linearization of SMF-based intensity-modulation/direct-detection systems. The convergence rate and power penalty are improved by a factor 2~3 (80km-SMF).