基于RLS_DGD的查找表更新算法

在间接学习结构的数字预失真技术中，需要使用自适应算法来更新查找表（LUT，LookUp Table）。在各种自适应算法中，RLS算法收敛快但是复杂度高。为降低其计算复杂度，这里提出采用二维坐标下降的RLS（RLS_DCD, RLS using Dichotomous Coordinate Descent）算法更新查找表，达到在数字预失真器系数的预测过程中以较低的运算量实现快速收敛的目的。仿真结果和分析表明，与采用RLS的自适应查找表更新算法比较，在邻信道功率比（ACPR, Adjacent Channel Power Ratio）改善指标相当的情况下，采用RLS—DCD的自适应查找表更新算法能大幅度降低其运算量。

Look-up Table Update Algorithm based on RLS_DCD

For the digital predistortion technology of indirect learning structure an adaptive algorithm is adopted to update the look-up table. Among those adaptive algorithms, RLS algorithm is fast in convergence but complex in computing. To reduce this computational complexity, RLS_DCD （RLS using dichotomous coordinate descent） algorithm for look-up table update is proposed, thus realizing fast convergence with fairly low computational complexity in the prediction of predistortion coefficients. Simulation and analysis results indicate that as compared with RLS-based look-up table update algorithm, the proposed algorithm could greatly reduce the computational complexity under the same condition of improved ACPR （Adjacent Channel Power Ratio）.