Characterizing prediction errors of a new tree height model for cut-to-length Pinus radiata stems through the Burr TypeⅫdistribution

Unlike height-diameter equations for standing trees commonly used in forest resources modelling,tree height models for cut-to-length(CTL)stems tend to produce prediction errors whose distributions are not conditionally normal but are rather leptokurtic and heavy-tailed.This feature was merely noticed in previous studies but never thoroughly investigated.This study characterized the prediction error distribution of a newly developed such tree height model for Pin us radiata(D.Don)through the three-parameter Burr TypeⅫ(BⅫ)distribution.The model’s prediction errors(ε)exhibited heteroskedasticity conditional mainly on the small end relative diameter of the top log and also on DBH to a minor extent.Structured serial correlations were also present in the data.A total of 14 candidate weighting functions were compared to select the best two for weightingεin order to reduce its conditional heteroskedasticity.The weighted prediction errors(εw)were shifted by a constant to the positive range supported by the BXII distribution.Then the distribution of weighted and shifted prediction errors(εw+)was characterized by the BⅫdistribution using maximum likelihood estimation through 1000 times of repeated random sampling,fitting and goodness-of-fit testing,each time by randomly taking only one observation from each tree to circumvent the potential adverse impact of serial correlation in the data on parameter estimation and inferences.The nonparametric two sample Kolmogorov-Smirnov(KS)goodness-of-fit test and its closely related Kuiper’s(KU)test showed the fitted BⅫdistributions provided a good fit to the highly leptokurtic and heavy-tailed distribution ofε.Random samples generated from the fitted BⅫdistributions ofεw+derived from using the best two weighting functions,when back-shifted and unweighted,exhibited distributions that were,in about97 and 95%of the 1000 cases respectively,not statistically different from the distribution ofε.Our results for cut-tolength P.radiata stems represented the first case of any tree species where a non-normal error distribution in tree height prediction was described by an underlying probability distribution.The fitted BXII prediction error distribution will help to unlock the full potential of the new tree height model in forest resources modelling of P.radiata plantations,particularly when uncertainty assessments,statistical inferences and error propagations are needed in research and practical applications through harvester data analytics.

融合IMR-WGAN的时序数据修复方法

工业数据由于技术故障和人为因素通常导致数据异常,现有基于约束的方法因约束阈值设置的过于宽松或严格会导致修复错误,基于统计的方法因平滑修复机制导致对时间步长较远的异常值修复准确度较低.针对上述问题,提出了基于奖励机制的最小迭代修复和改进WGAN混合模型的时序数据修复方法.首先,在预处理阶段,保留异常数据,进行信息标注等处理,从而充分挖掘异常值与真实值之间的特征约束.其次,在噪声模块提出了近邻参数裁剪规则,用于修正最小迭代修复公式生成的噪声向量.将其传递至模拟分布模块的生成器中,同时设计了一个动态时间注意力网络层,用于提取时序特征权重并与门控循环单元串联组合捕捉不同步长的特征依赖,并引入递归多步预测原理共同提升模型的表达能力;在判别器中设计了Abnormal and Truth奖励机制和Weighted Mean Square Error损失函数共同反向优化生成器修复数据的细节和质量.最后,在公开数据集和真实数据集上的实验结果表明,该方法的修复准确度与模型稳定性显著优于现有方法.

2.5D induced polarization forward modeling using the adaptive finite-element method

The conventional finite-element（FE） method often uses a structured mesh, which is designed according to the user’s experience, and it is not sufficiently accurate and flexible to accommodate complex structures such as dipping interfaces and rough topography. We present an adaptive FE method for 2.5D forward modeling of induced polarization（IP）. In the presented method, an unstructured triangulation mesh that allows for local mesh refinement and flexible description of arbitrary model geometries is used. Furthermore, the mesh refinement process is guided by dual error estimate weighting to bias the refinement towards elements that affect the solution at the receiver locations. After the final mesh is generated, the Jacobian matrix is used to obtain the IP response on 2D structure models. We validate the adaptive FE algorithm using a vertical contact model. The validation shows that the elements near the receivers are highly refined and the average relative error of the potentials converges to 0.4 % and 1.2 % for the IP response. This suggests that the numerical solution of the adaptive FE algorithm converges to an accurate solution with the refined mesh. Finally, the accuracy and flexibility of the adaptive FE procedure are also validated using more complex models.

Deep Unfolding for Cooperative Rate Splitting Multiple Access in Hybrid Satellite Terrestrial Networks

Rate splitting multiple access(RSMA)has shown great potentials for the next generation communication systems.In this work,we consider a two-user system in hybrid satellite terrestrial network(HSTN)where one of them is heavily shadowed and the other uses cooperative RSMA to improve the transmission quality.The non-convex weighted sum rate(WSR)problem formulated based on this model is usually optimized by computational burdened weighted minimum mean square error(WMMSE)algorithm.We propose to apply deep unfolding to solve the optimization problem,which maps WMMSE iterations into a layer-wise network and could achieve better performance within limited iterations.We also incorporate momentum accelerated projection gradient descent(PGD)algorithm to circumvent the complicated operations in WMMSE that are not amenable for unfolding and mapping.The momentum and step size in deep unfolding network are selected as trainable parameters for training.As shown in the simulation results,deep unfolding scheme has WSR and convergence speed advantages over original WMMSE algorithm.

Log Anomaly Detection Through GPT-2 for Large Scale Systems

As the scale of software systems expands,maintaining their stable operation has become an extraordinary challenge.System logs are semi-structured text generated by the recording function in the source code and have important research significance in software service anomaly detection.Existing log anomaly detection methods mainly focus on the statistical characteristics of logs,making it difficult to distinguish the semantic differences between normal and abnormal logs,and performing poorly on real-world industrial log data.In this paper,we propose an unsupervised framework for log anomaly detection based on generative pre-training-2(GPT-2).We apply our approach to two industrial systems.The experimental results on two datasets show that our approach outperforms state-of-the-art approaches for log anomaly detection.

Hybrid Architecture and Beamforming Optimization for Millimeter Wave Systems

Hybrid beamforming(HBF)has become an attractive and important technology in massive multiple-input multiple-output(MIMO)millimeter-wave(mmWave)systems.There are different hybrid architectures in HBF depending on different connection strategies of the phase shifter network between antennas and radio frequency chains.This paper investigates HBF optimization with different hybrid architectures in broadband point-to-point mmWave MIMO systems.The joint hybrid architecture and beamforming optimization problem is divided into two sub-problems.First,we transform the spectral efficiency maximization problem into an equivalent weighted mean squared error minimization problem,and propose an algorithm based on the manifold optimization method for the hybrid beamformer with a fixed hybrid architecture.The overlapped subarray architecture which balances well between hardware costs and system performance is investigated.We further propose an algorithm to dynamically partition antenna subarrays and combine it with the HBF optimization algorithm.Simulation results are presented to demonstrate the performance improvement of our proposed algorithms.

Using Moving North Pacific Index to Improve Rainy Season Rainfall Forecast over the Yangtze River Basin by Analog Error Correction

A new analog error correction （AEC） scheme based on the moving North Pacific index （MNPI） is designed in this study. This scheme shows obvious improvement in the prediction skill of the operational coupled general circulation model （CGCM） of the National Climate Center of China for the rainy season rainfall （RSR） anomaly pattern correlation coefficient （ACC） over the mid-to-lower reaches of the Yangtze River （MLRYR）. A comparative analysis indicates that the effectiveness of the new scheme using the MNPI is better than the system error correction scheme using the North Pacific index （NPI）. A Euclidean distance- weighted mean rather than a traditional arithmetic mean, is applied to the integration of the analog year＇s prediction error fields. By using the MNPI AEC scheme, independent sample hindcasts of RSR during the period 2003-2009 are then evaluated. The results show that the new scheme exhibited a higher forecast skill during 2003-2009, with an average ACC of 0.47; while the ACC for the NPI case was only 0.19. Furthermore, the forecast skill of the RSR over the MLRYR is examined. In the MNPI case, empirical orthogonal function （EOF） was used in the degree compression of the prediction error fields from the CCCM, whereas the AEC scheme was applied only to its first several EOF components for which the accumulative explained variance accounted for 80% of the total variance. This further improved the ACC of the independent sample hindcasts to 0.55 during the 7-yr period.

Ergodic secrecy capacity of MRC/SC in single-input multiple-output wiretap systems with imperfect channel state information

This paper investigates the secrecy performance of maximal ratio combining （MRC） and selection combining （SC） with imperfect channel state information （CSI） in the physical layer. In a single-input multiple- output （SIMO） wiretap channel, a source transmits confidential messages to the destination equipped with M antennas using the MRC/SC scheme to process the received multiple signals. An eavesdropper equipped with N antennas also adopts the MRC/SC scheme to promote successful eavesdropping. We derive the exact and asymptotic closed-form expressions for the ergodic secrecy capacity （ESC） in two cases： （1） MRC with weighting errors, and （2） SC with outdated CSI. Moreover, two important indicators, namely high signal-to-noise ratio （SNR） slope and high SNR power offset, which govern ESC at the high SNR region, are derived. Finally, simulations are conducted to validate the accuracy of our proposed analytical models. Results indicate that ESC rises with the increase of the number of antennas and the received SNR at the destination, and fades with the increase of those at the eavesdropper. Another finding is that the high SNR slope is constant, while the high SNR power offset is correlated with the number of antennas at both the destination and the eavesdropper.