Neural Networks for Financial Time Series Forecasting

Financial and economic time series forecasting has never been an easy task due to its sensibility to political, economic and social factors. For this reason, people who invest in financial markets and currency exchange are usually looking for robust models that can ensure them to maximize their profile and minimize their losses as much as possible. Fortunately, recently, various studies have speculated that a special type of Artificial Neural Networks (ANNs) called Recurrent Neural Networks (RNNs) could improve the predictive accuracy of the behavior of the financial data over time. This paper aims to forecast: (i) the closing price of eight stock market indexes; and (ii) the closing price of six currency exchange rates related to the USD, using the RNNs model and its variants: the Long Short-Term Memory (LSTM) and the Gated Recurrent Unit (GRU). The results show that the GRU gives the overall best results, especially for the univariate out-of-sample forecasting for the currency exchange rates and multivariate out-of-sample forecasting for the stock market indexes.     

Currency crises and the evolution of foreign exchange market: Evidence from minimum spanning tree

We examined the time series properties of the foreign exchange market for 1990-2008 in relation to the history of the currency crises using the minimum spanning tree (MST) approach and made several meaningful observations about the MST of currencies. First, around currency crises, the mean correlation coefficient between currencies decreased whereas the normalized tree length increased. The mean correlation coefficient dropped dramatically passing through the Asian crisis and remained at the lowered level after that. Second, the Euro and the US dollar showed a strong negative correlation after 1997, implying that the prices of the two currencies moved in opposite directions. Third, we observed that Asian countries and Latin American countries moved away from the cluster center (USA) passing through the Asian crisis and Argentine crisis, respectively.     

尖楔前体飞行器FADS系统的径向基网络建模及验证

文章研究了针对一种用于尖楔外形的嵌入式大气数据传感（flush air data sensing，FADS）系统的解算模型及精度.首先基于飞行包络及CFD数据建立了FADS系统的测压孔选取标准；然后基于径向基函数（radial basis function，RBF）的人工神经网络建模技术构建了FADS系统的网络解算模型；最后给出了模型的测试误差，分析了气动延时效应、位置误差等误差源模型对算法精度的影响，并给出了网络模型的预测精度.结果表明，针对尖楔外形测压孔配置特征，基于RBF的人工神经网络算法解算精度较好，攻角、侧滑角、Mach数及静压的网络输出预测值与真实值吻合较好，输出的测试误差（绝对值）分别小于0.25°，0.5°，0.05及250 Pa.结果同时表明神经网络建模技术在尖楔前体飞行器FADS系统中的有效性.      

神经群模型中癫痫状棘波的闭环控制性能研究

神经群模型可模拟产生癫痫发作间歇期、发作前期和发作期的脑电信号.本文基于代数估计法,给出一种新型的闭环反馈控制策略以消除神经群模型中的癫痫状棘波.代数估计法用以观测模型中的状态以进一步构造控制器.在多个神经群耦合的模型中,通过数值仿真研究了与所给的闭环反馈控制策略相关的一些特性,包括受控神经群的类型与消除棘波的能力之间的关系、受控神经群的数目与控制能量之间的关系、模型的参量和控制能量之间的关系,以期建立合适的控制规则实现利用尽可能小的控制能量消除癫痫状棘波.此外,通过数值仿真对基于代数估计法的闭环反馈控制策略和直接比例反馈控制策略进行比较,结果表明,利用代数估计法进行滤波能减少消除癫痫状棘波所需的控制能量.     

光折变神经网络

本文阐述了几种神经网络模型和模式两分法中感知算法的一种光学执行过程。利用光折变介质动态全息术来存入和修正光学神经网络的加权矩阵。该系统可用以完成模式分类。还介绍了光折变晶体LiNbO₃的一些实验结果。     

相关测速声呐阵元坐标估计算法研究

针对相关声呐接收阵元实际坐标与设计坐标不一致引起的测速误差,提出阵元坐标估计算法,以提高相关声呐的测速准确度。基于混响时空相关函数模型,对相关测速声呐时空相关函数特性进行了理论分析,提出了选择用延时0处的数据空间相关函数与理论空间相关函数进行阵元坐标估计的局域最小二乘拟合算法,并采用序列二次规划法解决了阵元坐标估计算法的最优化问题。海上试验数据的处理结果表明,该算法可以对阵元坐标进行有效估计,使用估计的阵元坐标后,相关测速声呐测速准确度提高了67%。      

Dynamics of electricity market correlations

Electricity market participants rely on demand and price forecasts to decide their bidding strategies, allocate assets, negotiate bilateral contracts, hedge risks, and plan facility investments. However, forecasting is hampered by the non-linear and stochastic nature of price time series. Diverse modeling strategies, from neural networks to traditional transfer functions, have been explored. These approaches are based on the assumption that price series contain correlations that can be exploited for model-based prediction purposes. While many works have been devoted to the demand and price modeling, a limited number of reports on the nature and dynamics of electricity market correlations are available. This paper uses detrended fluctuation analysis to study correlations in the demand and price time series and takes the Australian market as a case study. The results show the existence of correlations in both demand and prices over three orders of magnitude in time ranging from hours to months. However, the Hurst exponent is not constant over time, and its time evolution was computed over a subsample moving window of 250 observations. The computations, also made for two Canadian markets, show that the correlations present important fluctuations over a seasonal one-year cycle. Interestingly, non-linearities (measured in terms of a multifractality index) and reduced price predictability are found for the June-July periods, while the converse behavior is displayed during the December-January period. In terms of forecasting models, our results suggest that non-linear recursive models should be considered for accurate day-ahead price estimation. On the other hand, linear models seem to suffice for demand forecasting purposes.     

Multipath time delay estimation of underwater acoustic sinusoidal signals

To overcome the performance limitation of multipath time-delay estimation posed by underwater acoustic sinusoidal signals, an approach incorporating the frequency-domain weighting of the highly oscillatory Nonlinear Least Squares （NLS） cost function with the evolu- tionary optimization was proposed to facilitate the accurate estimation of the multipath timedelay of sinusoidal signals. In the described method, the number of the effective multipath signals, which is included into the parameter model as well as the multipath time-delay and amplitude factor, can be estimated simultaneously thus avoiding the requirement of additional computation. The experimental results performed with numerical simulation and sea-trial data are provided, demonstrating the effectiveness and precision enhancement of the proposed algorithm.     

Prediction of aqueous solubility of compounds based on neural network

Neural network algorithms are gradually applied to the field of chemical informatics. In this paper, the neural network models are used to predict the properties of compounds based on Molecular Information. We predict the aqueous solubility of compounds, and evaluate the prediction results of the Neural Networks including CNN, RNN, DNN, SNN. The performance of the models in predicting the solubility is able to meet or exceed the predicted effect of the method based on the molecular structure (ESOL). DNN model performance is of more accuracy, and RNN performance is of better stability. This method can directly avoid complex molecular structure characterisation, and provide a convenient and flexible way to predict properties of compounds.     

室内两步法监督式学习双耳声源距离估计

提出一种室内环境下两步法监督式学习双耳声源距离估计算法,该算法通过预先估计声源方位角信息以克服声源方位角的变化对声源距离估计性能的不利影响.该算法第1步利用深度神经网络模型估计声源的方位角,并将不同方位角的双耳信号分类;第2步中对每个方位角的双耳信号采用独立的深度神经网络模型进行声源距离估计,其中距离特征选用双耳信号的一些双耳特征和统计特性。在仿真和实际环境下,本文提出的两步法声源距离估计算法的距离估计准确率比现有算法提高了3%~5%左右,并且在各种不匹配环境下的距离估计准确率比现有算法高出5%~10%左右。实验结果表明利用声源方位角信息可以有效提高双耳声源距离估计算法的性能。      

A Nonlinear Maximum Correntropy Information Filter for High-Dimensional Neural Decoding

Neural signal decoding is a critical technology in brain machine interface (BMI) to interpret movement intention from multi-neural activity collected from paralyzed patients. As a commonly-used decoding algorithm, the Kalman filter is often applied to derive the movement states from high-dimensional neural firing observation. However, its performance is limited and less effective for noisy nonlinear neural systems with high-dimensional measurements. In this paper, we propose a nonlinear maximum correntropy information filter, aiming at better state estimation in the filtering process for a noisy high-dimensional measurement system. We reconstruct the measurement model between the high-dimensional measurements and low-dimensional states using the neural network, and derive the state estimation using the correntropy criterion to cope with the non-Gaussian noise and eliminate large initial uncertainty. Moreover, analyses of convergence and robustness are given. The effectiveness of the proposed algorithm is evaluated by applying it on multiple segments of neural spiking data from two rats to interpret the movement states when the subjects perform a two-lever discrimination task. Our results demonstrate better and more robust state estimation performance when compared with other filters.     

Design of Nonlinear Autoregressive Exogenous Model Based Intelligence Computing for Efficient State Estimation of Underwater Passive Target

In this study, an intelligent computing paradigm built on a nonlinear autoregressive exogenous (NARX) feedback neural network model with the strength of deep learning is presented for accurate state estimation of an underwater passive target. In underwater scenarios, real-time motion parameters of passive objects are usually extracted with nonlinear filtering techniques. In filtering algorithms, nonlinear passive measurements are associated with linear kinetics of the target, governing by state space methodology. To improve tracking accuracy, effective feature estimation and minimizing position error of dynamic passive objects, the strength of NARX based supervised learning is exploited. Dynamic artificial neural networks, which contain tapped delay lines, are suitable for predicting the future state of the underwater passive object. Neural networks-based intelligence computing is effectively applied for estimating the real-time actual state of a passive moving object, which follows a semi-curved path. Performance analysis of NARX based neural networks is evaluated for six different scenarios of standard deviation of white Gaussian measurement noise by following bearings only tracking phenomena. Root mean square error between estimated and real position of the passive target in rectangular coordinates is computed for evaluating the worth of the proposed NARX feedback neural network scheme. The Monte Carlo simulations are conducted and the results certify the capability of the intelligence computing over conventional nonlinear filtering algorithms such as spherical radial cubature Kalman filter and unscented Kalman filter for given state estimation model.     

A novel observer design method for neural mass models

Neural mass models can simulate the generation of electroencephalography(EEG) signals with different rhythms,and therefore the observation of the states of these models plays a significant role in brain research. The structure of neural mass models is special in that they can be expressed as Lurie systems. The developed techniques in Lurie system theory are applicable to these models. We here provide a new observer design method for neural mass models by transforming these models and the corresponding error systems into nonlinear systems with Lurie form. The purpose is to establish appropriate conditions which ensure the convergence of the estimation error. The effectiveness of the proposed method is illustrated by numerical simulations.     

近红外光谱与烟草样品总糖含量的非线性模型研究

针对烟草样品的近红外 (NIR)光谱与其总糖含量非线性相关的特点 ,提出了一种混合算法用于建立近红外光谱的非线性模型。该算法结合了偏最小二乘法 (PartialLeastSquare ,PLS)算法和人工神经网络(ArtificialNeuralNetwork ,ANN) ,把模型分成两个部分 :线性部分与非线性部分 ,并分别进行建模。与传统的多元校正算法PLS ,主成分回归 (PrincipleComponentRegression ,PCR) ,非线性PLS(NonlinearPLS ,NPLS)等相比 ,该混合算法所建的非线性参数模型的预测结果有明显的改善 ,从而为建立非线性模型提供了一种快速、准确的算法 ,可用于烟草样品总糖含量的定量分析。     

基于荧光光谱法与深度信念网络的稻种发芽率检测方法研究

针对传统稻种发芽率检测效率低、精度差、专业化要求高等问题,通过荧光光谱法结合深度信念网络(DBN)建立稻种发芽率预测模型。首先,将连粳7号和武运粳均分别老化0~7 d后,以5 min为间隔在纯净水中分别浸泡5~30 min。然后用荧光光谱仪检测浸泡液的荧光光谱,光谱数据经中心化后用集合经验模态分解(EEMD)去噪,并通过主成分分析法提取441.5 nm的特征荧光波长。最后,利用偏最小二乘回归(PLSR),反向传播神经网络(BPNN),径向基函数神经网络(RBFNN)和深度信念网络(DBN)建立水稻种子发芽预测模型。比较后得出,DBN模型在少数据、弱信号情况下的预测精度最高,预测集相关系数Rp和均方根误差RMSEP最大可达0.979 2和0.101。同时,通过分析混合稻种荧光数据R_p的变化趋势,得到最佳浸泡时间为22.1 min,实际上,精确度超过0.95(R_p)需要5 min左右。研究结果表明,结合荧光光谱法和EEMD-DBN模型,非破坏性地预测水稻种子发芽率具有可行性和高准确性,并且适用于不同颜色和污染水平的水稻种子的检测。     

SVM approach for predicting LogP

Summary The logarithm of the partition coefficient between n-octanol and water (logP) is an important parameter for drug discovery. Based upon the comparison of several prediction logP models, i.e. Support Vector Machines (SVM), Partial Least Squares (PLS) and Multiple Linear Regression (MLR), the authors reported SVM model is the best one in this paper.     

Decimative subspace-based parameter estimation methods of magnetic resonance spectroscopy based on prior knowledge

The method Hankel Total Least Squares (HTLS)-PK, which successfully incorporates prior knowledge of known signal poles into the method HTLS, has been proven to greatly improve the performance for parameter estimation of overlapping peaks of magnetic resonance spectroscopy (MRS) signal. In addition, decimation is also proposed as a way to increase the performance of subspace-based parameter estimation methods in the case of oversampling. Taking advantage of decimation in combination with prior knowledge to estimate the MRS signal parameters, two novel subspace-based parameter estimation methods, HTLSDSumPK and HTLSDStackPK, are presented in this paper. The experimental results and relevant analysis show that the methods HTLSDSumPK, HTLSDStackPK and HTLS-PK are slightly better than the method HTLS at low noise levels; however, the three prior-knowledge-incorporating methods, especially the method HTLSDSumPK, have much better performance than the method HTLS at high noise levels in the terms of robustness, estimated accuracy and computational complexity. Even if some inaccuracy of prior knowledge is considered, the method HTLSDSumPK also shows some advantages.     

基于向前和向后间隔偏最小二乘的特征光谱选择方法

在近红外光谱分析中,向前间隔偏最小二乘法（FiPLS）和向后间隔偏最小二乘法（BiPLS）是常用的基于波长变量选择的建模方法,其模型精度较高,但贪婪搜索特性较强,导致选出的波段并不能较好地反映待测成分的信息。针对该问题,提出一种基于两者组合策略的光谱特征波段选择方法（FB-iPLS）。在光谱分段的基础上,既利用FiPLS选取有用波段,同时利用BiPLS删除无用波段,来交互执行特征变量的选择与删除,对目标特征波段进行双向选择,用于提高模型的稳健性。用该方法建立水中乙醇含量的定量预测模型,并与FiPLS和BiPLS算法对比。由于光谱分段大小会对模型的结果有影响,该实验还考查这三种方法在不同光谱分段处的结果。在光谱划分60段时,提出的FB-iPLS方法取得最佳预测性能,其校正集与验证集相关系数r分别为0.967 7,0.967 0,交互验证均方根误差RMSECV分别为0.088 8,0.057 1。与FiPLS和BiPLS相比,该方法无论在不同光谱分段区间还是在各自最优与最差分段处,模型的整体预测性能都有所提高。实验结果表明,提出的方法能改善BiPLS与FiPLS贪婪搜索的特性,对特征波段的选取更高效、更具代表性,能进一步提高模型的预测性能。     

Long term memory in extreme returns of financial time series

It is well known that while daily price returns of financial markets are uncorrelated, their absolute values (‘volatility’) are long-term correlated. Here we provide evidence that certain subsequences of the returns themselves also exhibit long-term memory. These subsequences consist of maxima (or minima) of returns in consecutive time windows of R days. Our analysis shows that for both stocks and currency exchange rates, long-term correlations are significant for R≥4. We argue that this long-term memory which is similar to that observed in volatility clustering sheds further insight on price dynamics that might be used for risk estimation.     

Fast neutron spectra determination by threshold activation detectors using neural networks

Neural network method was used for fast neutron spectra unfolding in spectrometry by threshold activation detectors. The input layer of the neural networks consisted of 11 neurons for the specific activities of neutron-induced nuclear reaction products, while the output layers were fast neutron spectra which had been subdivided into 6, 8, 10, 12, 15 and 20 energy bins. Neural network training was performed by 437 fast neutron spectra and corresponding threshold activation detector readings. The trained neural network have been applied for unfolding 50 spectra, which were not in training sets and the results were compared with real spectra and unfolded spectra by SANDII. The best results belong to 10 energy bin spectra. The neural network was also trained by detector readings with 5% uncertainty and the response of the trained neural network to detector readings with 5%, 10%, 15%, 20%, 25% and 50% uncertainty was compared with real spectra. Neural network algorithm, in comparison with other unfolding methods, is very fast and needless to detector response matrix and any prior information about spectra and also the outputs have low sensitivity to uncertainty in the activity measurements. The results show that the neural network algorithm is useful when a fast response is required with reasonable accuracy.