基于误差模型的权重二值神经网络近似加速

针对智能识别系统精确度和硬件复杂度之间的均衡设计问题,提出了一种基于误差统计模型的权重二值神经网络近似加速方法。在提出了一种获得高精度轻量神经网络的权重二值化处理算法基础上,引入近似加法器、消除乘法器以进一步提高能效。最终提出了一种系统级误差统计模型用于系统评估和优化设计,该设计能够分析和预测权重二值神经网络近似加速系统的最终精度。结果表明:该模型可以准确地预测系统精度,与仿真结果对比,相对误差在2.05%~3.07%。该模型预测用于指导相应软硬件的设计优化,可大幅提高设计的迭代速度。

海洋环境数值预报检验方法的探讨与测试

对获得的海浪有效波高、海表温度、剖面温度和水位等要素的实测数据以及历史CTD资料,通过对这些数据进行分析和处理,检验WW3海浪、三维温盐流以及MODAS(模块化数据同化系统)等模式的数值预报产品,并设计相应的软件进行各时次的实时比对和旬月统计,自动生成客观的旬月检验报告。为改进预报模式、提高数值预报的准确性和预报产品的使用率提供可靠的依据。

一种顾及观测质量信息的自适应抗差Kalman滤波方法

针对观测数据含有异常粗差且无多余观测的应用情形,提出一种顾及观测质量信息的自适应抗差Kalman滤波方法。该方法两步计算得到自适应因子和抗差等价权矩阵,即首先利用顾及观测质量信息的抗差Kalman滤波得到消除观测粗差影响的参数估计值,然后根据该值构造动力学模型误差判别统计量并计算自适应因子。以某边坡GPS变形监测数据序列处理为例,利用RPDOP(Relative Position Dilution of Precision)值作为观测质量信息进行处理分析,结果表明该方法能够有效控制动力学模型误差和观测粗差对滤波估值的影响。

Online process monitoring for complex systems with dynamic weighted principal component analysis

Conventional multivariate statistical methods for process monitoring may not be suitable for dynamic processes since they usually rely on assumptions such as time invariance or uncorrelation. We are therefore motivated to propose a new monitoring method by compensating the principal component analysis with a weight approach.The proposed monitor consists of two tiers. The first tier uses the principal component analysis method to extract cross-correlation structure among process data, expressed by independent components. The second tier estimates auto-correlation structure among the extracted components as auto-regressive models. It is therefore named a dynamic weighted principal component analysis with hybrid correlation structure. The essential of the proposed method is to incorporate a weight approach into principal component analysis to construct two new subspaces, namely the important component subspace and the residual subspace, and two new statistics are defined to monitor them respectively. Through computing the weight values upon a new observation, the proposed method increases the weights along directions of components that have large estimation errors while reduces the influences of other directions. The rationale behind comes from the observations that the fault information is associated with online estimation errors of auto-regressive models. The proposed monitoring method is exemplified by the Tennessee Eastman process. The monitoring results show that the proposed method outperforms conventional principal component analysis, dynamic principal component analysis and dynamic latent variable.

Production response to price risk and market liberalization of Nigerian major agricultural crops

This study models supply response for major agricultural crops in Nigeria which include the standard arguments and price risk. The data comes from Central Bank of Nigeria annual reports and statement of account, National Bureau of Statistics＇ abstract of statistics and annual Agricultural survey manual. The data are analyzed using autoregressive distributed lag and cointegration and error correction models. The results indicate that producers are responsive not only to price but also to price risk and exchange rate.

TESTING SERIAL CORRELATION IN SEMIPARAMETRIC VARYING COEFFICIENT PARTIALLY LINEAR ERRORS-IN-VARIABLES MODEL

The authors propose a V_(N,p) test statistic for testing finite-order serial correlation in asemiparametric varying coefficient partially linear errors-in-variables model.The test statistic is shownto have asymptotic normal distribution under the null hypothesis of no serial correlation.Some MonteCarlo experiments are conducted to examine the finite sample performance of the proposed V_(N,p) teststatistic.Simulation results confirm that the proposed test performs satisfactorily in estimated sizeand power.

Residual Cusum Test for Parameters Change in ARCH Errors Models with Deterministic Trend

This paper analyzes the problem of testing for parameters change in ARCH errors models with deterministic trend based on residual cusum test. It is shown that the asymptotically limiting distribution of the residual cusum test statistic is still the sup of a standard Brownian bridge under null hypothesis. In order to check this, we carry out a Monte Carlo simulation and examine the return of IBM data. The results from both simulation and real data analysis support our claim. We also can explain this phenomenon from a theoretical viewpoint that the variance in ARCH model in mainly determined by its parameters.