ID 原文 译文
56288 充分发挥三值光学处理器位数众多、运算功能可重构、按位可分配等优势,设计出高效并行MSD (modified signed digit)数除法器对提高大数据除法的运算效率、促进三值光学计算机(ternary optical computer, TOC)在数值计算领域的应用意义重大. It is of greatsignificance to give full play to the advantages of ternary optical computer processors, such as a large numberof bits, reconfigurable operation functions, bit allocation, etc. by designing the efficient parallel MSD numberdivider to improve the division operation efficiency of large data and promote the application of ternary opticalcomputer in the field of numerical calculation.
56289 本文首次提出MSD数的符号判定算法,并基于SRT算法首次提出利用一个并行无进位SJ-MSD加法器和一个MSD数比较器实现单组MSD整数除法或多组MSD整数并行除法方案——并行MSD整数除法,该算法对于被除数等长的多组与单组MSD整数除法需要的机器周期是相同的. In this paper, the symbol decision algorithm for MSD numbers isproposed first time, and based on the SRT algorithm, the scheme for implementing a parallel division of multipleMSD integers using a parallel carry free SJ-MSD adder and an MSD comparator, which is called as parallelMSD integer division, is designed first time.
56290 实验表明,并行MSD整数除法方案是可行的,它将有效地提高大数据处理效率并加速TOC进入数值计算等实际应用领域. For the corresponding parallel MSD integer divider, the machinecycles required to run multiple MSD integer division in parallel is the same as the machine cycles required for theoperation of a single MSD integer division. Experiments show that the parallel SRT integer division scheme isfeasible. It will effectively improve the efficiency of large data processing and accelerate TOC to enter practicalapplications such as numerical computation.
56291 针对池计算网络的构建问题,提出了一种稀疏连接的异步神经元池构造方法,该方法将多个子神经元池顺序连接,并在子神经元池之间设置滞后环节,以实现各子神经元池对输入信息的异步处理,进而构成串行的记忆. In order to solve the reservoir computing network construction problem, a sparsely connected asyn?chronous neuron reservoir construction method is proposed. The method connects several sub-reservoirs sequen?tially and sets lag links among sub-reservoirs in order to handle input signals asynchronously in sub-reservoirs,and further constitutes serial memory.
56292 为实现信息高效传输,子神经元池之间采用稀疏的连接方式. In order to achieve efficient information transmission, sparse connectionsare used among sub-reservoirs.
56293 实验表明,所提方法能够有效地提高神经元池的记忆容量,易于解决长时依赖问题. Experimental results show that the proposed method can effectively improvethe memory capacity of reservoir and it is easy to deal with long-term dependence problems.
56294 此外,该结构能够使神经元池产生丰富的动力学行为,对初始参数也有较好的鲁棒性. In addition, theproposed structure makes the reservoir produce more abundant dynamic behavior and has better robustness tothe initial parameters.
56295 社交网络蕴含着丰富的多媒体信息,如何实现社交网络跨媒体信息的搜索已成为研究热点. Social networks contain abundant multimedia information. The research of cross-media informationsearch from social networks has become a hotspot.
56296 基于深度学习的单一模态语义特征提取和学习在社交网络信息搜索上取得了较好的效果. Single-modal semantic feature extractions and learnings basedon deep learning show an appropriate effect on information search from social networks.
56297 在跨模态信息搜索时不同模态的数据特征不能直接比较,因此不同模态之间的语义鸿沟是亟待解决的关键问题. But features of differentmodalities cannot be directly compared, therefore the semantic gap between different modalities is a key problemto be resolved for cross-media information search.