Joint Classification and Regression for Visual Tracking with Fully Convolutional Siamese Networks

International Journal of Computer Vision - Visual tracking of generic objects is one of the fundamental but challenging problems in computer vision. Here, we propose a novel fully convolutional...     

A modified teaching–learning-based optimisation algorithm for bi-objective re-entrant hybrid flowshop scheduling

In this paper, a modified teaching–learning-based optimisation (mTLBO) algorithm is proposed to solve the re-entrant hybrid flowshop scheduling problem (RHFSP) with the makespan and the total tardiness criteria. Based on the simple job-based representation, a novel decoding method named equivalent due date-based permutation schedule is proposed to transfer an individual to a feasible schedule. At each generation, a number of superior individuals are selected as the teachers by the Pareto-based ranking phase. To enhance the exploitation ability in the promising area, the insertion-based local search is embedded in the search framework as the training phase for the TLBO. Due to the characteristics of the permutation-based discrete optimisation, the linear order crossover operator and the swap operator are adopted to imitate the interactions among the individuals in both the teaching phase and the learning phase. To store the non-dominated solutions explored during the search process, an external archive is used and updated when necessary. The influence of the parameter setting on the mTLBO in solving the RHFSP is investigated, and numerical tests with some benchmarking instances are carried out. The comparative results show that the proposed mTLBO outperforms the existing algorithms significantly.     

Algorithm for the extraction of page-formatted binary digital data

We present an algorithm for fast and reliable extraction of page-formatted binary digital data. The advantages of the algorithm include a low raw bit-error rate, fast extraction speed, the use of a simple and density-efficient coding scheme, and large tolerance to a change of the signal-to-noise ratio. We used this algorithm to analyze shot-noise-limited binary data that had large interpage and intrapage intensity variations and obtained an improvement in the bit-error rate of 3-4 orders of magnitude compared with that in a single-threshold-detection scheme. Implications of our results for the development of high-speed, high-density holographic memories are discussed.     

The first measurement of an absolute surface concentration of reaction intermediates in ethylene hydrogenation

The first measurement of a turnover rate with respect to surface intermediate concentration in a high pressure heterogeneous catalytic reaction is reported. By using infrared-visible sum frequency generation to study the hydrogenation of ethylene on Pt(111), it was found that the surface concentration of -bonded ethylene, the key reaction intermediate, represented approximately 4% of a monolayer. Thus the absolute turnover rate per surface adsorbed ethylene molecule is 25 times faster than the rate measured per platinum atom. To explain these results, we propose a model of weakly adsorbed ethylene intermediates reacting on atop sites.     

Load-balanced parallel streamline generation on large scale vector fields

Because of the ever increasing size of output data from scientific simulations, supercomputers are increasingly relied upon to generate visualizations. One use of supercomputers is to generate field lines from large scale flow fields. When generating field lines in parallel, the vector field is generally decomposed into blocks, which are then assigned to processors. Since various regions of the vector field can have different flow complexity, processors will require varying amounts of computation time to trace their particles, causing load imbalance, and thus limiting the performance speedup. To achieve load-balanced streamline generation, we propose a workload-aware partitioning algorithm to decompose the vector field into partitions with near equal workloads. Since actual workloads are unknown beforehand, we propose a workload estimation algorithm to predict the workload in the local vector field. A graph-based representation of the vector field is employed to generate these estimates. Once the workloads have been estimated, our partitioning algorithm is hierarchically applied to distribute the workload to all partitions. We examine the performance of our workload estimation and workload-aware partitioning algorithm in several timings studies, which demonstrates that by employing these methods, better scalability can be achieved with little overhead.     

Optimal Error Estimates of the Legendre Tau Method for Second-Order Differential Equations

In this paper, we prove that the Legendre tau method has the optimal rate of convergence in L ²-norm, H ¹-norm and H ²-norm for one-dimensional second-order steady differential equations with three kinds of boundary conditions and in C([0,T];L ²(I))-norm for the corresponding evolution equation with the Dirichlet boundary condition. For the generalized Burgers equation, we develop a Legendre tau-Chebyshev collocation method, which can also be optimally convergent in C([0,T];L ²(I))-norm. Finally, we give some numerical examples.     

A system-level model for a silicon thermal flow sensor

A system-level model with lumped parameters for a thermal flow sensor is presented. The model is built with 13 circuit cells consisting of thermal resistors and thermal capacitors in SPICE. The circuit cell originates from the heat conduction equation using the Finite Differential Method, including the 2-D thermal conduction cell, the convection cell, and the thermal capacity in the chip. Based on the thermal model of the flow sensor, the 2-D temperature distribution of the chip can be calculated with SPICE in both the constant power mode (CP) and constant temperature difference mode (CTD). As an example, the system level model of the thermal anemometer in the CTD mode was established in PSPICE. Wind tunnel test was carried out to verify the system model, and show a reasonable agreement with the simulation results, with an error less than 8%.     

VLCC中的DAG并行算法

基于组件的密码学虚拟实验室（VLCC）采用有向无环图（DAG）的拓扑排序机制管理组件。在分析VLCC各组件之间的数据依赖和运行次序关系的基础上,提出一种新的基于Java多线程机制和“唤醒”机制的DAG并行算法。与拓扑排序算法相比,具有低算法时间复杂度的特点。实验结果表明,系统在新算法下较大地缩短了系统运行时间,提高资源使用效率和用户满意度,能更好地完善VLCC。