ID 原文 译文
19055 在此基础上，用相关熵损失函数替代均方误差(MSE)损失函数，推导出用于训练F-ELM各层权重矩阵的相关熵循环更新公式，以增强其分类能力与鲁棒性。 On this basis, the Mean Square Error (MSE) loss function isreplaced by the correntropy-based loss function. A correntropy-based cycle update formula for training theweight matrices of the F-ELM is derived to enhance classification ability and robustness.
19056 为了检验方法的可行性，该文分别在数据库Caltech 101, MSRC和15 Scene上进行实验。 Extensive experimentsare performed on Caltech 101, MSRC and 15 Scene datasets respectively.
19057 实验结果证明，该文所提CF-ELM能够在原有基础上进一步融合表示级特征，从而提高分类正确率。 The experimental results show thatCF-ELM can further fuse the representation level features to improve the classification accuracy.
19058 提高时频分辨率对多分量非平稳信号的分析与重建具有至关重要的作用。 The improvement of time-frequency resolution plays a crucial role in the analysis and reconstructionof multi-component non-stationary signals.
19059 传统的时频分析方法由于窗口固定，分析频率变化较快的信号时存在时频聚集性不高的问题，无法自适应分辨多分量信号。 For traditional time-frequency analysis methods with fixed window, the time-frequency concentration is low and hardly to distinguish the multi-component signals with fast-varying frequencies.
19060 该文针对频率快速变化信号，利用信号的局部信息特征，提出一种自适应的时频同步压缩变换算法。 In this paper, by adopting the local information of the signal, an adaptive synchro squeezing transform is proposed for the signals with fast-varying frequencies.
19061 该方法有效提升了已有同步压缩变换时频分辨率，特别适用于频率接近且快速变换的多分量信号。 The proposed method is with high time-frequency resolution, superior to existing synchrosqueezing methods, and particularly suitable for multi-component signals with close and fast-varying frequencies.
19062 同时，利用可分性条件，该文提出利用局部瑞利熵值对自适应窗口参数进行估计。 Meanwhile, by using the separability condition, the adaptive window parameters are estimated by local Rényi entropy.
19063 最后，通过对合成信号和实测信号分析，证明了所提方法的可行性，对分析和重建复杂非平稳信号具有重要意义。 Finally, experiments on synthetic and real signals demonstrate the correctness of the proposed method, which is suitable to analyze and recover complex non-stationary signals.
19064 基于粗糙集理论的粗糙熵阈值法不需要图像之外的先验信息。 Image thresholding methods based on the rough entropy segment the images without prior information except the images.