Nianyin Zeng

A survey of deep neural network architectures and their applications

This work was supported in part the Royal Society of the UK, the National Natural Science Foundation of China under Grants 61329301, 61374010, and 61403319, and the Alexander von Humboldt Foundation of Germany.

A Hybrid EKF and Switching PSO Algorithm for Joint State and Parameter Estimation of Lateral Flow Immunoassay Models

In this paper, a hybrid extended Kalman filter (EKF) and switching particle swarm optimization (SPSO) algorithm is proposed for jointly estimating both the parameters and states of the lateral flow immunoassay model through available short time-series measurement. Our proposed method generalizes the well-known EKF algorithm by imposing physical constraints on the system states. Note that the state constraints are encountered very often in practice that give rise to considerable difficulties in system analysis and design. The main purpose of this paper is to handle the dynamic modeling problem with state constraints by combining the extended Kalman filtering and constrained optimization algorithms via the maximization probability method. More specifically, a recently developed SPSO algorithm is used to cope with the constrained optimization problem by converting it into an unconstrained optimization one through adding a penalty term to the objective function. The proposed algorithm is then employed to simultaneously identify the parameters and states of a lateral flow immunoassay model. It is shown that the proposed algorithm gives much improved performance over the traditional EKF method.

Deep Belief Networks for Quantitative Analysis of a Gold Immunochromatographic Strip

Gold immunochromatographic strip (GICS) has become a popular membrane-based diagnostic tool in a variety of settings due to its sensitivity, simplicity and rapidness. This paper aimed to develop a framework of automatic image inspection to further improve the sensitivity as well as the quantitative performance of the GICS systems. As one of the latest methodologies in machine learning, the deep belief network (DBN) is applied, for the first time, to quantitative analysis of GICS images with hope to segment the test and control lines with a high accuracy. It is remarkable that the exploited DBN is capable of simultaneously learning three proposed features including intensity, distance and difference to distinguish the test and control lines from the region of interest that are obtained by preprocessing the GICS images. Several indices are proposed to evaluate the proposed method. The experiment results show the feasibility and effectiveness of the DBN in the sense that it provides a robust image processing methodology for quantitative analysis of GICS.

A Novel Switching Delayed PSO Algorithm for Estimating Unknown Parameters of Lateral Flow Immunoassay

In this paper, the parameter identification problem of the lateral flow immunoassay (LFIA) devices is investigated via a new switching delayed particle swarm optimization (SDPSO) algorithm. By evaluating an evolutionary factor in each generation, the velocity of the particle can adaptively adjust the model according to a Markov chain in the proposed SDPSO method. During the iteration process, the SDPSO can adaptively select the inertia weight, acceleration coefficients, locally best particle pbest and globally best particle gbest in the swarm. It is worth highlighting that the pbest and the gbest can be randomly selected from the corresponding values in the previous iteration. That is, the delayed information of the pbest and the gbest can be exploited to update the particle’s velocity in current iteration according to the evolutionary states. The strategy can not only improve the global search but also enhance the possibility of eventually reaching the gbest. The superiority of the proposed SDPSO is evaluated on a series of unimodal and multimodal benchmark functions. Results demonstrate that the novel SDPSO algorithm outperforms some well-known PSO algorithms in aspects of global search and efficiency of convergence. Finally, the novel SDPSO is successfully exploited to estimate the unknown time-delay parameters of a class of nonlinear state-space LFIA model.

A switching delayed PSO optimized extreme learning machine for short-term load forecasting☆

Abstract In this paper, a hybrid learning approach, which combines the extreme learning machine (ELM) with a new switching delayed PSO (SDPSO) algorithm, is proposed for the problem of the short-term load forecasting (STLF). In particular, the input weights and biases of ELM are optimized by a new developed SDPSO algorithm, where the delayed information of locally best particle and globally best particle are exploited to update the velocity of particle. By testing the proposed SDPSO-ELM in a comprehensive manner on a tanh function, this approach obtain better generalization performance and can also avoid adding unnecessary hidden nodes and overtraining problems. Moreover, it has shown outstanding performance than other state-of-the-art ELMs. Finally, the proposed SDPSO-ELM algorithm is successfully applied to the STLF of power system. Experiment results demonstrate that the proposed learning algorithm can get better forecasting results in comparison with the radial basis function neural network (RBFNN) algorithm.

Inferring nonlinear lateral flow immunoassay state-space models via an unscented Kalman filter

This paper is concerned with the problem of learning structure of the lateral flow immunoassay (LFIA) devices via short but available time series of the experiment measurement. The model for the LFIA is considered as a nonlinear state-space model that includes equations describing both the biochemical reaction process of LFIA system and the observation output. Especially, the time-delays occurring among the biochemical reactions are considered in the established model. Furthermore, we utilize the unscented Kalman filter (UKF) algorithm to simultaneously identify not only the states but also the parameters of the improved state-space model by using short but high-dimensional experiment data in terms of images. It is shown via experiment results that the UKF approach is particularly suitable for modelling the LFIA devices. The identified model with time-delay is of great significance for the quantitative analysis of LFIA in both the accurate prediction of the dynamic process of the concentration distribution of the antigens/antibodies and the performance optimization of the LFIA devices.概要创新点1) 本文建立的非线性状态空间模型由生化反应系统方程和观测方程组成. 特别地, 模型还考虑了存在于各个反应中的时滞现象. 2) 基于获取的短时间序列数据, 无迹卡尔曼滤波能够同时准确辨识出模型中的状态及参数. 3) 该模型可观察和预测试条的免疫反应动态特性, 并能够辅助优化试条的定量特性.

Inference of Nonlinear State-Space Models for Sandwich-Type Lateral Flow Immunoassay Using Extended Kalman Filtering

In this paper, a mathematical model for sandwich-type lateral flow immunoassay is developed via short available time series. A nonlinear dynamic stochastic model is considered that consists of the biochemical reaction system equations and the observation equation. After specifying the model structure, we apply the extended Kalman filter (EKF) algorithm for identifying both the states and parameters of the nonlinear state-space model. It is shown that the EKF algorithm can accurately identify the parameters and also predict the system states in the nonlinear dynamic stochastic model through an iterative procedure by using a small number of observations. The identified mathematical model provides a powerful tool for testing the system hypotheses and also for inspecting the effects from various design parameters in both rapid and inexpensive way. Furthermore, by means of the established model, the dynamic changes in the concentration of antigens and antibodies can be predicted, thereby making it possible for us to analyze, optimize, and design the properties of lateral flow immunoassay devices.

Image-Based Quantitative Analysis of Gold Immunochromatographic Strip via Cellular Neural Network Approach

Gold immunochromatographic strip assay provides a rapid, simple, single-copy and on-site way to detect the presence or absence of the target analyte. This paper aims to develop a method for accurately segmenting the test line and control line of the gold immunochromatographic strip (GICS) image for quantitatively determining the trace concentrations in the specimen, which can lead to more functional information than the traditional qualitative or semi-quantitative strip assay. The canny operator as well as the mathematical morphology method is used to detect and extract the GICS reading-window. Then, the test line and control line of the GICS reading-window are segmented by the cellular neural network (CNN) algorithm, where the template parameters of the CNN are designed by the switching particle swarm optimization (SPSO) algorithm for improving the performance of the CNN. It is shown that the SPSO-based CNN offers a robust method for accurately segmenting the test and control lines, and therefore serves as a novel image methodology for the interpretation of GICS. Furthermore, quantitative comparison is carried out among four algorithms in terms of the peak signal-to-noise ratio. It is concluded that the proposed CNN algorithm gives higher accuracy and the CNN is capable of parallelism and analog very-large-scale integration implementation within a remarkably efficient time.

Facial expression recognition via learning deep sparse autoencoders

Abstract Facial expression recognition is an important research issue in the pattern recognition field. In this paper, we intend to present a novel framework for facial expression recognition to automatically distinguish the expressions with high accuracy. Especially, a high-dimensional feature composed by the combination of the facial geometric and appearance features is introduced to the facial expression recognition due to its containing the accurate and comprehensive information of emotions. Furthermore, the deep sparse autoencoders (DSAE) are established to recognize the facial expressions with high accuracy by learning robust and discriminative features from the data. The experiment results indicate that the presented framework can achieve a high recognition accuracy of 95.79% on the extended Cohn–Kanade (CK+) database for seven facial expressions, which outperforms the other three state-of-the-art methods by as much as 3.17%, 4.09% and 7.41%, respectively. In particular, the presented approach is also applied to recognize eight facial expressions (including the neutral) and it provides a satisfactory recognition accuracy, which successfully demonstrates the feasibility and effectiveness of the approach in this paper.

Identification of Nonlinear Lateral Flow Immunoassay State-Space Models via Particle Filter Approach

In this paper, the particle filtering approach is used, together with the kernel smoothing method, to identify the state-space model for the lateral flow immunoassay through available but short time-series measurement. The lateral flow immunoassay model is viewed as a nonlinear dynamic stochastic model consisting of the equations for the biochemical reaction system as well as the measurement output. The renowned extended Kalman filter is chosen as the importance density of the particle filter for the purpose of modeling the nonlinear lateral flow immunoassay. By using the developed particle filter, both the states and parameters of the nonlinear state-space model can be identified simultaneously. The identified model is of fundamental significance for the development of lateral flow immunoassay quantification. It is shown that the proposed particle filtering approach works well for modeling the lateral flow immunoassay.