Transfer learning allows leveraging the knowledge of source domains, available a priori, to help training classifier for target domain, where data is scarce. The effectiveness transfer affected by relationship between and target. Rather than improving learning, brute force poorly related may decrease performance. One strategy reduce this negative import from multiple sources increase chance finding one closely This work extends boosting framework transferring sources. Two new algorithms,...

We propose novel Stacked Spatio-Temporal Graph Convolutional Networks (Stacked-STGCN) for action segmentation, i.e., predicting and localizing a sequence of actions over long videos. extend the Network (STGCN) originally proposed skeleton-based recognition to enable nodes with different characteristics (e.g., scene, actor, object, action), feature descriptors varied lengths, arbitrary temporal edge connections account large graph deformation commonly associated complex activities. further...

Digital imaging systems with extreme zoom capabilities are traditionally found in astronomy and wild life monitoring. More recently, the need for such has extended to long range surveillance wide area monitoring as forest fires, airport perimeters, harbors, waterways. Auto-focusing is an indispensable function designed applications. This paper studies feasibility of image based passive auto-focusing control high magnification on off-the-shelf telescopes digital cameras/camcorders,...

The MapReduce framework has become the de facto scheme for scalable semi-structured and un-structured data processing in recent years. Hadoop ecosystem evolved into its second generation, YARN, which adopts fine-grained resource management schemes job scheduling. One of primary performance concerns YARN is how to minimize total completion length, i.e., makespan, a set jobs. However, precedence constraint or fairness current widely used scheduling policies such as FIFO Fair, can both lead...

The carbon emission trading market is an important policy tool to promote the realization of China’s peaking and neutrality goals. Research on relationship between other related ones supports formulation risk aversion. Firstly, we construct Carbon–Energy–Stock system compare information spillover three subsystems under a unified framework. Secondly, adopt connectedness network identify role status carbon, energy, stock markets. Thirdly, through rolling window approach, explore dynamic...

Dual-camera systems have been widely used in surveillance because of the ability to explore wide field view (FOV) omnidirectional camera and zoom range PTZ camera. Most existing algorithms require a priori knowledge camera's projection model solve nonlinear spatial correspondences between two cameras. To overcome this limitation, methods are proposed: 1) geometry 2) homography calibration, where polynomials with automated selection approximate mapping, respectively. The proposed not only...

Most existing camera placement algorithms focus on coverage and/or visibility analysis, which ensures that the object of interest is visible in camera's field view (FOV). However, visibility, a fundamental requirement tracking, insufficient for automated persistent surveillance. In such applications, continuous consistently labeled trajectory same should be maintained across different views. Therefore, sufficient uniform overlap between cameras' FOVs secured so handoff can successfully and...

Hadoop is an emerging framework for parallel big data processing. While becoming popular, too complex regular users to fully understand all the system parameters and tune them appropriately. Especially when processing a batch of jobs, default setting may cause inefficient resource utilization unnecessarily prolong execution time. This paper considers extremely important slot configuration which by fixed static. We proposed enhanced called FRESH can derive best setting, dynamically configure...

The MapReduce framework has become the defacto scheme for scalable semi-structured and un-structured data processing in recent years. Hadoop ecosystem evolved into its second generation, YARN, which adopts fine-grained resource management schemes job scheduling. Nowadays, fairness efficiency are two main concerns YARN because resources shared contended by multiple applications. However, current scheduling does not yield optimal arrangement, unnecessarily causing idle inefficient It omits...

Recent studies have shown that combining Transformer and conditional strategies to deal with offline reinforcement learning can bring better results. However, in a conventional scenario, the agent receive single frame of observations one by according its natural chronological sequence, but Transformer, series are received at each step. Individual features cannot be extracted efficiently make more accurate decisions, it is still difficult generalize effectively for data outside distribution....

The MapReduce framework and its open source implementation Hadoop have become the defacto platform for scalable analysis on large data sets in recent years. One of primary concerns is how to minimize completion length (i.e., makespan) a set jobs. current only allows static slot configuration, i.e., fixed numbers map slots reduce throughout lifetime cluster. However, we found that such configuration may lead low system resource utilizations as well long length. Motivated by this, propose...

To better understand the characteristics of a bike-sharing system, we applied complex network methods to analyze relationship between stations within system. Firstly, using Gephi software, constructed public bicycle networks different urban areas based on real-time data Nanjing Secondly, analyzed and compared degree, strength, radiation distance, community structure internal relations The results showed that there were many with low usage bicycles. Furthermore, was geographical division...

Convolutional neural networks (CNNs) are widely used in computer vision and natural language processing. Field-programmable gate arrays (FPGAs) popular accelerators for CNNs. However, if critical applications, the reliability of FPGA-based CNNs becomes a priority because FPGAs prone to suffer soft errors. Traditional protection schemes, such as triple modular redundancy (TMR), introduce large overhead, which is not acceptable resource-limited platforms. This article proposes use an ensemble...

In recent years, tourism has become a significant driver of many countries’ economies. To maximize revenue from tourism, it is crucial to prioritize the effective management scenic spots and tourist attractions, also raise awareness about these places. Social media platforms have played pivotal role in promoting as users frequently share videos reviews related tourism. Analyzing managing essential for understanding tourists’ opinions specific destinations. this study, we evaluated spot by...

This study posits that soil particles in the filter layer are ellipsoidal. The effective pore radius of material was calculated for various particle-shape parameters and distributions. relationship between porosity ratio long axis to short ellipsoidal a loose arrangement also examined. results indicate initially decreases subsequently increases with axis; however, rate increase progressively slows. A method transforming irregularly shaped into forms is proposed. parameter, S, introduced...

The MapReduce paradigm and its open source implementation Hadoop are emerging as an important standard for large-scale data-intensive processing in both industry academia. A cluster is typically shared among multiple users with different types of workloads. When a flock jobs concurrently submitted to cluster, they compete the resources overall system performance terms job response times, might be seriously degraded. Therefore, one challenging issue ability efficient scheduling such...

Most existing camera placement algorithms focus on coverage and/or visibility analysis, which ensures that the object of interest is visible in camera—s field view (FOV). However, visibility, a fundamental requirement tracking, insufficient for persistent and automated tracking. In such applications, continuous consistently labeled trajectory same should be maintained across different cameras' views. Therefore, sufficient overlap between FOVs secured so handoff can executed successfully...

In order to achieve improved recognition performance in comparison with conventional broadband images, this paper addresses a new method that automatically specifies the optimal spectral range for multispectral face images according given illuminations. The novelty of our lies introduction distribution separation measure and selection by ranking these values. selected ranges are consistent physics analysis imaging process. fused from chosen verified outperform 3%-20%, based on variety...

Existing calibration algorithms address the problem of covariate shift via unsupervised domain adaptation. However, these methods suffer from following limitations: 1) they require unlabeled data target domain, which may not be available at stage in real-world applications and 2) their performance depends heavily on disparity between distributions source domains. To two limitations, we present novel solutions generalization. Our core idea is to leverage multiple domains reduce effective...