ID	原文	译文
19365	其次，设计了基于4并行快速滤波算法的卷积计算加速单元，该单元采用若干小滤波器组成的复杂度较低的并行滤波结构来实现。	Secondly, a convolution calculation accelerating unit based on 4parallel fast filtering algorithm is designed. The unit is realized by a less complex parallel filtering structure composed of several small filters.
19366	利用手写数字集MNIST对所设计的CNN加速器电路进行测试，结果表明：在xilinx kintex7平台上，输入时钟为100 MHz时，电路的计算性能达到了20.49 GOPS，识别率为98.68%。	Using the handwritten digit set MNIST to test the designed CNN accelerator circuit, the results show that on the xilinx kintex7 platform, when the input clock is 100 MHz, the computational performance of the circuit reaches 20.49 GOPS, and the recognition rate is 98.68%.
19367	可见通过减少CNN的计算量，能够提高电路的计算性能。	It can be seen that the computational performance of the circuit can be improved by reducing the amount of calculation of the CNN.
19368	码流生成在FPGA电子设计自动化(EDA)流程中，提供应用电路在芯片上物理实现所需的精准配置信息。	Bitstream generator in FPGA Electronic Design Automation(EDA) offers precise configuration information, which enables the application circuits to be implemented on the target device.
19369	现代FPGA的发展一方面呈现出器件规模及码流容量越来越大的趋势，另一方面越来越多可变阵列大小的嵌入式应用(例如eFPGA)又要求码流生成器具备更高的配置效率以及更精简的可重构数据库。	On one hand, modern FPGAs tend to have larger device scale and more configuration bits, on the other hand, embedded applications (e.g. eFPGAs) require better configuration efficiency and smaller, more adaptive database.
19370	针对码流生成时间增加的问题和阵列规模任意缩放的需求，该文提出一种模式匹配和层次映射的码流生成方法，即对编程单元按配置模式进行分类建模，在配置时按模型进行调用匹配，并采用了层次化的码流映射策略，使得数据库可随阵列排布调整动态生成。	In order to meet these new requirements, a bit-stream generation method is proposed which firstly models the configurable resources by configuration modes and matches the netlist with these models, then hierarchical mapping strategy is used to search every bit on a dynamically generated database determined by the array floor plan.
19371	该方法可有效应对FPGA嵌入式应用中码流容量的增大以及阵列规模可变所带来的挑战，同时相比平面化的建模及映射方法，码流配置的时间复杂度由O(n)降低为O(lgn)。	This method well meets the challenges that embedded applications may bring-the surge ofconfiguration bit count and the changeable size of the array. Compared to flattened modelling and mappingmethod, its time complexity is reduced from O(n) to O(lgn).
19372	针对条件对抗领域适应(CDAN)方法未能充分挖掘样本的可迁移性，仍然存在部分难以迁移的源域样本扰乱目标域数据分布的问题，该文提出一种基于迁移权重的条件对抗领域适应(TW-CDAN)方法。	Considering the failure of the Conditional adversarial Domain AdaptatioN(CDAN) to fully utilize thesample transferability, which still struggle with some hard-to-transfer source samples disturbed the distributionof the target domain samples, a Transfer Weight based Conditional adversarial Domain AdaptatioN(TW-CDAN) is proposed.
19373	首先利用领域判别模型的判别结果作为衡量样本迁移性能的主要度量指标，使不同的样本具有不同的迁移性能；	Firstly, the discriminant results in the domain discriminant model as the main factor areemployed to measure the transfer performance.
19374	其次将样本的可迁移性作为权重应用在分类损失和最小熵损失上，旨在消除条件对抗领域适应中难以迁移样本对模型造成的影响；	Then the weight is applied to class loss and minimum entropyloss. It is for eliminating the influence of hard-to-transfer samples of the model.