ID 原文 译文
47036 对动作区间上界和下界进行加权来求最优动作， Upper and the lower bounds ofthe action range were weighted to obtain the optimal action.
47037 然后通过线性函数逼近器来近似动作区间上下界的权值， The two bounds were approximated by linear function.
47038 将最优动作求解转换为对双策略参数向量的求解。 Af-terward, the problem of obtaining the optimal action was transferred to the learning of double policy parameter vectors.
47039 为了加快上下界的参数向量学习速率，设计了增量的 Fisher 信息矩阵和动作上下界权值的资格迹， To speed the learning, the incremental Fisher information matrix and the eligibilities of both bounds were designed.
47040 为了证明该算法的有效性，将该算法与其他连续动作空间的经典强化学习算法在 3 个强化学习的经典测试实验中进行比较。实验结果表明，所提算法具有收敛速度快和收敛稳定性好的优点。 Atthree reinforcement learning problems, compared with other representative methods with continuous action space, thesimulation results show that the proposed algorithm has the advantages of rapid convergence rate and high convergencestability.
47041 视网膜分层是视盘结构三维分析、青光眼三维特征提取的基础， Retina layering was the basis of optic disc structure analysis and 3D feature extraction of glaucoma.
47042 为改善视网膜 OCT 图像的层次分割效果，提出了一个基于强度的二维视网膜黄斑 OCT 图像多层结构分割算法。 In order toimprove the layering effect of retinal OCT images, an intensity based multilayer segmentation algorithm fortwo-dimensional retinal OCT images was proposed.
47043 分割方法通过预处理、滤波等操作，计算出视网膜 OCT 图像中每个 A-scan 的强度和强度梯度值，能很好地分割出视网膜 OCT 图像中的 RNFL 上界、IS 和 OS 分界线、RPE 层下界等， Through preprocessing and filtering operation, the segmentation al-gorithm calculated the intensity and intensity gradient of each A-scan in the retinal OCT image to obtain the upper boundRNFL, the dividing line IS and OS, and the lower bound RPE.
47044 并用最短距离计算的黄斑距离策略对黄斑部位分层结果进行再优化，从而实现视网膜 OCT 图像的层次分割。 Then macular distance strategy, calculated by the shortest dis-tance, was used to further optimize the layering result of macular area, so as to achieve layering segmentation of the retinal OCTimages.
47045 实验结果表明，所提算法优化效果好，时间复杂度低，运行速度较快。 The experimental results show the algorithm has good optimization effect, low time complexity and fast running speed.