ID 原文 译文
55447 在在线非线性自适应滤波应用中,由于基于多核学习的算法具有更高自由度并且能够利用更多数据特征,相比基于单核学习的算法在性能上有很大提升。 In online nonlinear adaptive filter applications, because based on multi-core learning algorithm has a higher degree of freedom and the ability to use more data characteristics, compared based on a single core learning algorithm has a lot to improve on the performance.
55448 首先给出具有相同"字典"的多核仿射投影算法,该算法是多核学习方法和仿射投影算法的结合。 First set with the same "dictionary" multicore affine projection algorithm, this algorithm is more nuclear and affine projection algorithm combined with learning method.
55449 然后基于相干准则针对多核仿射投影算法的特例,对应不同高斯核带宽,利用相干稀疏准则构造不同"字典",提出利用自适应l1-范数正则项来解决归一化多核最小均方非线性自适应滤波算法在非平稳信号下"字典"存在冗余核函数的问题。 Then based on the coherence criterion for the special case of multi-core affine projection algorithm, corresponding to different gaussian kernel bandwidth, using different coherence rule of sparse structure "dictionary", proposed by using adaptive l1 norm regularization - to solve the normalized multi-core minimum mean square nonlinear adaptive filter algorithm under non-stationary signal "dictionary" has the problem of redundant kernel function.
55450 最后数值仿真结果与比较验证了所提算法的有效性。 Finally the results of numerical simulation and compared to verify the effectiveness of the proposed algorithm.
55451 线性高斯状态空间模型中假设噪声为已知的白噪声过于苛刻。 A linear state space model hypothesis gaussian noise is known as the white noise is too harsh.
55452 认为过程噪声与观测噪声均未知且二者的解析关系确定,假设观测噪声的均值非零且服从高斯分布,方差服从逆威沙特分布,从而构成了层次式贝叶斯模型。 Think the process noise and observation noise are unknown and the analytical relationship of the two factors, average nonzero and of observation noise are assumed to be gaussian distribution, and inverse power Saudi distributed variance, which constitutes the hierarchical bayesian model.
55453 利用变分推断将均值与方差和系统状态一起作为随机变量进行迭代估计,在得到观测噪声的均值与方差的估计值后,利用其与过程噪声的关系进一步更新未知过程噪声的均值与方差,从而动态地得到每一时刻过程噪声与观测噪声的一、二阶统计矩信息,即使在噪声统计信息动态变化的情况下也有较满意的滤波精度。 Using variational inference will mean value and variance with state of the system as a random variable iteration estimates, in after get the estimates of the mean value and variance of noise, using its relationship with the process noise further update the unknown process mean and variance of noise, thus dynamically get every moment of process noise and observation noise, second-order statistical moment information, even in the case of noise statistical information dynamic change has satisfactory accuracy.
55454 实验证明了该算法的有效性。 Experimental results show the effectiveness of the algorithm.
55455 优化选择一定的行动策略能促使任务联盟向期望的目标效果演化。 Optimization to select certain strategies to task alliance evolution to the desired target effect.
55456 考虑部分事件/行动在不同时段下影响强度不相一致,使用考虑影响值时变的动态影响网对联盟演化过程行动策略优选问题进行建模,给出因果强度逻辑下概率传播参数设计的一致性条件,并基于因果强度逻辑进行影响值计算。 Considered part of the event/action under different times impact strength is not consistent, use considering influence values time-varying dynamic effect of the network optimization problem for alliance evolution strategies, probability in the strength of the causal logic are spreading the consistency condition of parameter design, and impact value is calculated based on the strength of the causal logic.