ID 原文 译文
17745 首先，针对增广状态的跳变马尔科夫系统，该文给出了联合估计增广状态和量测噪声协方差矩阵的变分贝叶斯推断概率模型。 Firstly, the variational Bayesian inference probabilistic model of theaugmented state and the covariance matrix of the measurement noise for the jump Markovarian system ispresented.
17746 其次，通过理论推导证明了该概率模型是非共轭的。 Secondly, the probabilistic model is proven to be non-conjugated.
17747 最后，通过引入一种“信息反馈+后处理”方案，提出联合后验密度的次优求解方法。 Finally, by introducing a novelpost processing method, the suboptimal solution to calculate the joint posterior distribution is proposed.
17748 所提算法能够在线估计未知的量测噪声协方差矩阵，具有更强的鲁棒性和适应性。 The proposed algorithm can estimate the unknown covariance matrix of the measurement noise online, thus it is more robust and has higher adaptability.
17749 仿真结果验证了算法的有效性。 Simulation result verifies good performance of the proposed algorithm.
17750 Adaboost是一种广泛使用的机器学习算法，然而Adaboost算法在训练时耗时十分严重。 The Adaboost algorithm provides noteworthy benefits over the traditional machine algorithms fornumerous applications, including face recognition, text recognition, and pedestrian detection. However, it takesa lot of time during the training process that affects the overall performance.
17751 针对该问题，该文提出一种基于自适应权值的Adaboost快速训练算法AWTAdaboost。 Adaboost fast training algorithmbased on adaptive weight (Adaptable Weight Trimming Adaboost, AWTAdaboost) is proposed in this work toaddress the aforementioned issue.
17752 该算法首先统计每一轮迭代的样本权值分布，再结合当前样本权值的最大值和样本集规模计算出裁剪系数，权值小于裁剪系数的样本将不参与训练，进而加快了训练速度。 First, the algorithm counts the current sample weight distribution of eachiteration. Then, it combines the maximum value of current sample weights with data size to calculate theadaptable coefficients. The sample whose weight is less than the adaptable coefficients is discarded, that speedsup the training.
17753 在INRIA数据集和自定义数据集上的实验表明，该文算法能在保证检测效果的情况下大幅加快训练速度， The experimental results validate that it can significantly speed up the training speed whileensuring the detection effect.
17754 相比于其他快速训练算法，在训练时间接近的情况下有更好的检测效果。 Compared with other fast training algorithms, the detection effect is better whenthe training time is close to each other.