基于鲁棒回声状态网络的混沌时间序列预测研究

针对回声状态网络模型易受异常点影响的问题, 提出一种基于拉普拉斯先验分布的鲁棒回声状态网络模型. 通过采用对异常点不敏感的拉普拉斯分布代替高斯分布作为模型输出的先验, 以增强网络对于异常点的抑制能力. 此外, 为解决由引入拉普拉斯分布所造成的求解困难的问题, 根据边际优化方法, 构建适当的替代函数, 使拉普拉斯先验等价转化为易于计算的高斯形式, 并通过贝叶斯方法实现模型参数的自适应估计. 仿真结果表明, 在异常点存在的情况下, 本文所提出的模型具有较好的鲁棒性, 并仍能保持较高的预测精度. 关键词： 回声状态网络 鲁棒模型 替代函数 拉普拉斯分布     

改进的贝叶斯压缩感知目标方位估计

针对基于高斯先验模型的贝叶斯压缩感知在目标方位(Direction Of Arrival,DOA)估计中可能出现明显随机伪峰的问题,改进了高斯先验模型,并在此基础上提出了一种贝叶斯压缩感知目标方位估计方法。通过波束输出噪声背景预估与二值指示变量标记,并引入基于信号先验方差的噪声方差估计方法,与变分贝叶斯推断相结合改进目标方位估计性能和优化迭代收敛过程。利用32元线阵对改进算法进行数值仿真处理和分析结果表明,该改进方法不仅可以准确估计目标信号的方位,而且可以显著地减少空间谱中伪峰的数量。实际海上实验数据处理结果表明,使用改进后的贝叶斯压缩感知方法进行DOA估计,可以显著地抑制空间谱中随机的伪峰,提高波束输出峰值背景比,具有更强的目标检测能力。      

基于逆散射理论的地震波速度正则化反演

本文基于逆散射理论利用正则化有限差分对比源反演算法对地震波传播速度进行反演, 该方法是基于波动方程的频率域波形反演算法, 利用非线性共轭梯度法, 通过最小化目标优化函数不断迭代更新速度模型. 由于地球物理反演问题的病态性和不稳定性, 通过基于反演参数总变差的正则化处理, 使反演问题变为良性问题且算法具有较强的抗噪声干扰能力. 反演过程中使用了频率-空间域9点差分正演算子以及PML吸收边界条件. 与其他反演算法相比, 由于背景模型在反演迭代过程中保持不变, 可以避免在每次迭代过程中重新构造正演算子及矩阵分解等相关计算过程, 使得该算法非常适合于大规模三维反演计算. 此外, 本文采用基于MPI的并行计算, 进一步提高了反演计算的效率. 二维CSEG模型反演结果表明该方法可以反演得到高分辨率的地震波速度重建结果, 为地震勘探数据处理及解释提供准确的速度信息.     

高平均功率电光开关晶体热畸变分析

 深入分析了方板构型的电光开关晶体在高功率载荷条件下的热畸变行为，讨论了光强分布对热效应的影响。以KDP晶体为例，分别计算了激光束光强为高斯分布和均匀分布时晶体的温升、相应的热应力分布、波前畸变以及热退偏。结果表明，光强的分布形式对波前畸变和热退偏的影响是不同的。相对于光强均匀分布的激光束，高斯光束减缓了光斑边沿处的温度梯度，产生的热应力较小，因此可以减弱热退偏效应；另一方面，在光束口径范围内，高斯光束产生了附加的温度分布非均匀性，因而波前畸变会大一些。     

基于广义M估计的鲁棒容积卡尔曼滤波目标跟踪算法

为减小测量异常误差对非线性目标跟踪系统的影响, 提出了一种基于广义M估计的鲁棒容积卡尔曼滤波算法. 首先将非线性测量方程等价变换, 利用约束总体最小二乘准则构建广义M估计极值函数, 在不进行线性化近似的前提下将其引入到容积卡尔曼滤波求解框架中. 然后根据Mahalanobis距离构建异常误差判别量, 利用卡方分布的置信水平确定判决门限, 并建立改进的三段Huber权函数, 使其能够降低小异常误差权值, 剔除大异常误差. 理论分析表明, 该方法具有无需求导、跟踪精度高、实时性好等优点, 且无需已知异常误差的统计特性; 实验结果表明, 所提算法能够有效减小异常误差的影响, 在实际非线性物理系统中具有广阔的应用空间.     

Influence of statistical distribution properties on ultrafast random-number generation using chaotic semiconductor lasers

We study the influence of statistical distribution properties on ultrafast random-number generators (RNGs) using chaotic laser system consisting of a semiconductor laser subject to dual-chaotic optical injections. Two completely different distributions are considered in this paper: one is a long-tailed distribution, and the other is a well-fitted Gaussian distribution. The numerical results show that, using minimum post-processing, symmetric distribution allows for the extraction of 4 least significant bits (LSBs) per sample; while for the asymmetric distribution the produced sequence based on the last LSB still exhibits certain bias. In other words, the important role of symmetric distribution in fast generation of random bits using multi-bit extraction scheme is demonstrated in numerical simulations.     

基于多轴差分吸收光谱技术测量对流层HCHO垂直分布

甲醛（HCHO）在大气光化学反应中扮演着重要的角色，是一种重要的气溶胶前体物和光化学氧化指示剂。大气中HCHO的来源主要是直接排放和光化学反应生成。大气光化学反应与太阳辐射强度密切相关，一般来说，太阳辐射强度越强，大气光化学反应越剧烈，HCHO的二次来源产率也就越高。故针对HCHO的研究成为当今大气环境研究的一个重要课题。介绍了基于多轴差分吸收光谱技术 (MAX-DOAS) 获取对流层HCHO垂直柱浓度（VCD）及垂直廓线的反演算法。该方法是基于非线性最优估算法的两步反演方法，首先反演气溶胶垂直廓线, 然后在此基础上反演HCHO垂直廓线。其中第二步气体廓线反演时，气溶胶廓线线型会影响气体廓线反演的权重函数从而影响气体垂直廓线反演的精度, 为此, 研究了三种不同气溶胶廓线类型（指数型、高斯型和玻尔兹曼型）对HCHO垂直廓线反演的影响。结果表明，在三种气溶胶廓线类型条件下，当气溶胶光学厚度(AOD)为0.1时，气体反演的总误差、平均核的包络线、灵敏高度上限、自由度以及HCHO垂直廓线结果都比较接近，即气溶胶廓线类型对HCHO垂直廓线反演的影响很小。而对于200 m以下（含200 m）的近地面，通过指数型、高斯型和玻尔兹曼型气溶胶廓线获取的HCHO体积混合比（VMR）与真实HCHO VMR的差异分别为36.89%，-0.04%和23.30%, 表明使用指数型和玻尔兹曼型气溶胶廓线类型反演HCHO垂直廓线会高估近地面HCHO浓度，而高斯型气溶胶廓线类型则正好相反。此外，还反演了北京国科大站点一次污染过程中HCHO的垂直廓线，分析了污染过程中HCHO的垂直分布特征。结果表明，HCHO主要集中在1.0 km以下且一天中高值出现在午后，主要来自于本地产生，即西南风将污染的VOCs气团带到观测点，经过本地的光化学反应产生HCHO而积累，造成了此次HCHO浓度升高。结合气流后向轨迹分析，来自站点西南方向的输送是引起HCHO污染的重要原因。故观测站点的HCHO主要受污染输送和二次氧化的影响。最后对比了此次污染过程中不同气溶胶条件对HCHO廓线反演的误差影响。结果显示，气溶胶浓度高时，反演的灵敏高度和自由度下降，反演的高度分辨率下降，且反演总误差增加。     

Automatic segmentation of white matter lesions on magnetic resonance images of the brain by using an outlier detection strategy

White matter lesions (WMLs) are commonly observed on the magnetic resonance (MR) images of normal elderly in association with vascular risk factors, such as hypertension or stroke. An accurate WML detection provides significant information for disease tracking, therapy evaluation, and normal aging research. In this article, we present an unsupervised WML segmentation method that uses Gaussian mixture model to describe the intensity distribution of the normal brain tissues and detects the WMLs as outliers to the normal brain tissue model based on extreme value theory. The detection of WMLs is performed by comparing the probability distribution function of a one-sided normal distribution and a Gumbel distribution, which is a specific extreme value distribution. The performance of the automatic segmentation is validated on synthetic and clinical MR images with regard to different imaging sequences and lesion loads. Results indicate that the segmentation method has a favorable accuracy competitive with other state-of-the-art WML segmentation methods.     

微粒特征分布函数测量理论和方法的进一步研究

运用微粒散射形体微扰理论对任意形体微粒的研究结果,根据光敏区光强的Gauss分布特征和微粒穿越位置的随机性,建立了相关模型和数学方程并求得了任意位置的微粒形体影响因子分布函数的数值解,讨论了大变化区的非线性影响.其次,根据微粒数分布离散性的特征,建立了微粒粒度分布函数的反演矩阵及其实验方法.最后进行了实验验证.     

Some optimizations on detecting gravitational wave using convolutional neural network

This work investigates the problem of detecting gravitational wave (GW) events based on simulated damped sinusoid signals contaminated with white Gaussian noise. It is treated as a classification problem with one class for the interesting events. The proposed scheme consists of the following two successive steps: decomposing the data using a wavelet packet, representing the GW signal and noise using the derived decomposition coefficients; and determining the existence of any GW event using a convolutional neural network (CNN) with a logistic regression output layer. The characteristic of this work is its comprehensive investigations on CNN structure, detection window width, data resolution, wavelet packet decomposition and detection window overlap scheme. Extensive simulation experiments show excellent performances for reliable detection of signals with a range of GW model parameters and signal-to-noise ratios. While we use a simple waveform model in this study, we expect the method to be particularly valuable when the potential GW shapes are too complex to be characterized with a template bank.     

Automatic segmentation of cerebral white matter hyperintensities using only 3D FLAIR images

Magnetic Resonance (MR) white matter hyperintensities have been shown to predict an increased risk of developing cognitive decline. However, their actual role in the conversion to dementia is still not fully understood. Automatic segmentation methods can help in the screening and monitoring of Mild Cognitive Impairment patients who take part in large population-based studies. Most existing segmentation approaches use multimodal MR images. However, multiple acquisitions represent a limitation in terms of both patient comfort and computational complexity of the algorithms. In this work, we propose an automatic lesion segmentation method that uses only three-dimensional fluid-attenuation inversion recovery (FLAIR) images. We use a modified context-sensitive Gaussian mixture model to determine voxel class probabilities, followed by correction of FLAIR artifacts. We evaluate the method against the manual segmentation performed by an experienced neuroradiologist and compare the results with other unimodal segmentation approaches. Finally, we apply our method to the segmentation of multiple sclerosis lesions by using a publicly available benchmark dataset. Results show a similar performance to other state-of-the-art multimodal methods, as well as to the human rater.     

基于鲁棒极端学习机的混沌时间序列建模预测

针对混沌时间序列预测模型易受异常点影响,导致模型预测精度低的问题,在贝叶斯框架下提出一种鲁棒极端学习机.所提模型将具有重尾分布特性的高斯混合分布作为模型输出似然函数,得到一种对异常点和噪声更具鲁棒性的预测模型.但由于将高斯混合分布作为模型输出似然函数后,模型输出的边缘似然函数变成难以解析处理的形式,因此引入变分方法进行近似推理,实现模型参数的估计.在加入异常点和噪声的情况下,将所提模型应用于大气环流模拟模型方程Lorenz序列以及Rossler混沌时间序列和太阳黑子混沌时间序列的预测中,预测结果验证了所提模型的有效性.     

核磁自旋回波串的液体分量分解快速反演法(英文)

该文叙述核磁自旋回波串的液体分量分解快速反演法.此方法假定液体,无论是在散装形式或饱和多孔介质中,可以用一个或一组核磁弛豫线形来表征.对一维核磁共振的拉普拉斯反演,它可以是预先确定的一个或一组T2或T1分布.对二维核磁共振的拉普拉斯反演,它可以是一个或一组预先确定的( D, T2)或( T1, T2)二维分布.对三维核磁共振的拉普拉斯反演,它可以是一个或一组预先设定的( D, T1, T2)三维分布.这些预先确定的线形,可以是高斯、B样条或预先由实验或经验确定的任何线形.这种方法可以显着降低核磁共振数据反演的计算时间,特别是从石油核磁共振测井采集的多维数据反演,它不需牺牲反演所得的分布的平滑性和准确性.这种方法的另一个新应用是作为一种约束求解方法来过滤相邻深度所采集的数据噪音.核磁共振测井的噪音信号,往往造成在相邻深度的同一岩性岩层有不同的T2分布.在此情况下, T2分布就不能用来识别岩性.通过非一般的矩阵操作,作者成功实现了对相邻深度的回波串实施约束求解方法,从而使得T2分布成为一种可靠的岩性识别指标.     

铷蒸气激光器3D理论模型的建立及阈值特性模拟

基于三能级速率方程理论,结合惠更斯-菲涅尔衍射积分公式,利用分步光束传播方法建立了半导体泵浦碱金属蒸气激光器(DPAL)端面泵浦3D理论计算模型,用于模拟研究铷蒸气激光器的泵浦阈值特性。该模型将光束传播、光与物质相互作用独立考虑,利用迭代算法求解。考虑激光器腔内模式与泵浦光模式匹配对阈值特性的影响,模拟了单端泵浦铷蒸气激光器内部的三维动力学过程。在具体算例中研究了模式匹配最佳时,在阈值工作状态下铷蒸气室中的三维粒子数分布,以及沿不同轴线的增益分布。仿真了在最佳模式匹配位置附近,泵浦光模式变化对阈值特性的影响。根据蒸气池中粒子数分布和光场分布具体分析了模式匹配影响阈值特性的机理。同时还模拟了不同长度蒸气池对阈值特性的影响。结果表明,腔参数与泵浦光模式之间存在一定的关系,对于特定的泵浦光模式,存在恰当的腔参数,使得阈值达到最小。     

Focusing properties of Gaussian beam by a nonspiral phase plate system

In this article, a nonspiral plate is investigated numerically by vector diffraction theory to observe the focusing properties of Gaussian beams. Both low- and high-numerical-aperture (NA) optical systems are considered in the investigation. It is found that the parameter of NA and the phase vary rate of the phase plate influence the focal intensity distribution considerably. When a nonspiral plate is used to provide linear phase variation on one half of the Gaussian beam, it may adjust the focal spot considerably and conveniently. Changing the vary rate of the phase plate or the parameter of NA can alter optical intensity distribution; some novel focal spots and focal switch may also occur.     

DOE effect on BER performance in MSK space uplink chaotic optical communication

With the increasing demand for space optical communication security, space chaotic optical communication has attracted a great amount of attention. Compared with traditional space optical communication, a chaotic optical communication system has a higher bit error rate(BER) for its complex system design. In order to decrease the BER of space chaotic optical communication systems, we introduce two diffractive optical elements(DOEs)at a transmitting terminal(Tx). That is because the commonly used reflective optical antenna at Tx blocks the central part of the transmission beam, which leads to a great amount of power consumption. Introducing the DOEs into the optical subsystem at Tx can reshape the transmission beam from a Gaussian distribution to a hollow Gaussian distribution so that the block of the secondary mirror in the reflective optical antenna can be avoided. In terms of the DOE influence on communication quality, we give a BER model based on a minimumshift-key(MSK) space uplink chaotic optical communication system to describe the DOE function. Based on the model, we further investigate the effect of the DOEs through analyzing the BER relationship versus basic system parameters based on the BER model. Both different mismatch conditions of chaotic systems and different atmospheric turbulence conditions are considered. These results will be helpful for the scheme design of space uplink chaotic optical communication systems.     

Active learning for OPM in FMF systems

Optical performance monitoring (OPM) is essential to guarantee the robust and reliable operation of few-mode fiber (FMF)-based transmission. The available OPM methods including the analytical models such as the enhanced Gaussian noise model provide high accuracy along with high computational complexity which makes them improper for real-time implementations. As an alternative approach, machine learning (ML)-based OPM removes this barrier at the cost of leveraging a large training dataset. However, generating a field or synthetic dataset for FMF-based transmission is very hard and time-consuming. As a specific ML deployment, active learning (AL) is designed to work with a small training dataset, therefore, in this paper, we employ AL for OPM in FMF-based transmission. Results indicate that the proposed AL-based OPM can properly estimate the generalized signal-to-noise ratio by using a very small training dataset and achieve the root mean squared error similar to that obtained by working on large training datasets.     

Approaching Adversarial Example Classification with Chaos Theory

Adversarial examples are one of the most intriguing topics in modern deep learning. Imperceptible perturbations to the input can fool robust models. In relation to this problem, attack and defense methods are being developed almost on a daily basis. In parallel, efforts are being made to simply pointing out when an input image is an adversarial example. This can help prevent potential issues, as the failure cases are easily recognizable by humans. The proposal in this work is to study how chaos theory methods can help distinguish adversarial examples from regular images. Our work is based on the assumption that deep networks behave as chaotic systems, and adversarial examples are the main manifestation of it (in the sense that a slight input variation produces a totally different output). In our experiments, we show that the Lyapunov exponents (an established measure of chaoticity), which have been recently proposed for classification of adversarial examples, are not robust to image processing transformations that alter image entropy. Furthermore, we show that entropy can complement Lyapunov exponents in such a way that the discriminating power is significantly enhanced. The proposed method achieves 65% to 100% accuracy detecting adversarials with a wide range of attacks (for example: CW, PGD, Spatial, HopSkip) for the MNIST dataset, with similar results when entropy-changing image processing methods (such as Equalization, Speckle and Gaussian noise) are applied. This is also corroborated with two other datasets, Fashion-MNIST and CIFAR 19. These results indicate that classifiers can enhance their robustness against the adversarial phenomenon, being applied in a wide variety of conditions that potentially matches real world cases and also other threatening scenarios.     

The improvement of cross-calibration of IIM data and band selection for FeO inversion

Chang’E-1 (CE-1) Interference Imaging Spectrometer (IIM) dataset suffers from the weak response in the near infrared (NIR) bands, which are the important wavelength for retrieving the minerals and elements of the Moon. In this paper, the cross-calibration was implemented to the IIM hyperspectral data for improving the weak response in NIR bands. The results show that the cross-calibrated IIM spectra were consistent to the Earth-based telescopic spectra, which suggests that the cross-calibration yields acceptable results. For further validating the influence of the cross-calibration on the FeO inversion and searching the optimal bands to retrieve lunar FeO contents, four band selection schemes were designed to retrieve FeO using the original and cross-calibrated IIM spectra. By comparing the distribution patterns and histograms of the IIM derived FeO contents with the Clementine derived FeO, the IIM 891 nm band after cross-calibration showed a higher accuracy in the FeO inversion, hence most useful for lunar FeO inversion.