An improved BP algorithm for pattern recognition

An improved BP algorithm for pattern recognition is proposed in this paper. By a function substitution for error measure, it resolves the inconsistency of BP algorithm for pattern recognition problems, i.e. the quadratic error is not sensitive to whether the training pattern is recognized correctly or not. Trained by this new method, the computer simulation result shows that the convergence speed is increased to treble and performance of the network is better than conventional BP algorithm with momentum and adaptive step size.

An Ameliorated Phase Generated Carrier Demodulation Algorithm With Low Harmonic Distortion and High Stability

A novel ameliorated phase generated carrier (PGC) demodulation algorithm based on arctangent function and differential-self-multiplying (DSM) is proposed in this paper. The harmonic distortion due to nonlinearity and the stability with light intensity disturbance (LID) are investigated both theoretically and experimentally. The nonlinearity of the PGC demodulation algorithm has been analyzed and an analytical expression of the total-harmonic-distortion (THD) has been derived. Experimental results have confirmed the low harmonic distortion of the ameliorated PGC algorithm as expected by the theoretical analysis. Compared with the traditional PGC-arctan and PGC-DCM algorithm, the ameliorated PGC algorithm has a much lower THD as well as a better signal-to-noise-and-distortion (SINAD). A THD of below 0.1% and a SINAD of 60 dB have been achieved with PGC modulation depth (value) ranges from 1.5 to 3.5 rad. The stability performance with LID has also been studied. The ameliorated PGC algorithm has a much higher stability than the PGC-DCM algorithm. It can keep stable operations with LID depth as large as 26.5 dB and LID frequency as high as 1 kHz. The system employing the ameliorated PGC demodulation algorithm has a minimum detectable phase shift of 5 mu rad/root Hz @ 1 kHz, a large dynamic range of 120 dB @ 100 Hz, and a high linearity of better than 99.99%.

missing data imputation: a fuzzy k-means clustering algorithm over sliding window

Tianjin University of Technology

Solving MIN UR problem by triangle evolution algorithm with archiving and niche techniques

In this paper, as an extension of minimum unsatisfied linear relations problem (MIN ULR), the minimum unsatisfied relations (MIN UR) problem is investigated. A triangle evolution algorithm with archiving and niche techniques is proposed for MIN UR problem. Different with algorithms in literature, it solves MIN problem directly, rather than transforming it into many sub-problems. The proposed algorithm is also applicable for the special case of MIN UR, in which it involves some mandatory relations. Numerical results show that the algorithm is effective for MIN UR problem and it outperforms Sadegh's algorithm in sense of the resulted minimum inconsistency number, even though the test problems are linear.

A finite-element algorithm for modeling variably saturated flows

A general numerical algorithm in the context of finite element scheme is developed to solve Richards’ equation, in which a mass-conservative, modified head based scheme (MHB) is proposed to approximate the governing equation, and mass-lumping techniques are used to keep the numerical simulation stable. The MHB scheme is compared with the modified Picard iteration scheme (MPI) in a ponding infiltration example. Although the MHB scheme is a little inferior to the MPI scheme in respect of mass balance, it is superior in convergence character and simplicity. Fully implicit, explicit and geometric average conductivity methods are performed and compared, the first one is superior in simulation accuracy and can use large time-step size, but the others are superior in iteration efficiency. The algorithm works well over a wide variety of problems, such as infiltration fronts, steady-state and transient water tables, and transient seepage faces, as demonstrated by its performance against published experimental data. The algorithm is presented in sufficient detail to facilitate its implementation.

基于克隆选择算法的非线性优化迭代学习控制；Clonal selection algorithm based nonlinear optimal iterative learning control

为实现对模型不确定的有约束非线性系统在特定时间域上输出轨迹的有效跟踪,将改进的克隆选择算法用于求解迭代学习控制中的优化问题。提出基于克隆选择算法的非线性优化迭代学习控制。在每次迭代运算后,一个克隆选择算法用于求解下次迭代运算中的最优输入,另一个克隆选择算法用于修正系统参考模型。仿真结果表明,该方法比GA-ILC具有更快的收敛速度,能够有效处理输入上的约束以及模型不确定问题,通过少数几次迭代学习就能取得满意的跟踪效果。