Update Model Parameters Research Articles

Video Extrapolation Method Based on Time-Varying Energy Optimization and CIP.

Video extrapolation/prediction methods are often used to synthesize new videos from images. For fluid-like images and dynamic textures as well as moving rigid objects, most state-of-the-art video extrapolation methods use non-physics-based models that learn orthogonal bases from a number of images but at high computation cost. Unfortunately, data truncation can cause image degradation, i.e., blur, artifact, and insufficient motion changes. To extrapolate videos that more strictly follow physical rules, this paper proposes a physics-based method that needs only a few images and is truncation-free. We utilize physics-based equations with image intensity and velocity: optical flow, Navier-Stokes, continuity, and advection equations. These allow us to use partial difference equations to deal with the local image feature changes. Image degradation during extrapolation is minimized by updating model parameters, where a novel time-varying energy balancer model that uses energy based image features, i.e., texture, velocity, and edge. Moreover, the advection equation is discretized by high-order constrained interpolation profile for lower quantization error than can be achieved by the previous finite difference method in long-term videos. Experiments show that the proposed energy based video extrapolation method outperforms the state-of-the-art video extrapolation methods in terms of image quality and computation cost.

A hierarchical elastic full-waveform inversion scheme based on wavefield separation and the multistep-length approach

Elastic full-waveform inversion (FWI) updates model parameters by minimizing the residuals of the P- and S-wavefields, resulting in more local minima and serious nonlinearity. In addition, the coupling of different parameters degrades the inversion results. To address these problems, we have developed a hierarchical elastic FWI scheme based on wavefield separation and a multistep-length gradient approach. First, we have derived the gradients expressed by different wave modes; analyzed the crosstalk between various parameters; and evaluated the sensitivity of separated P-wave, separated S-wave, and P- and S-wave misfit functions. Then, a practical four-stage inversion workflow was developed. In the first stage, conventional FWI is used to achieve rough estimates of the P- and S-wave velocities. In the second stage, we only invert the P-wave velocity applying the separated P-wavefields when strong S-wave energy is involved, or we merely update the S-wave velocity by matching the separated S-wavefields for the weak S-wave case. The PP and PS gradient formulas are used in these two cases, respectively. Therefore, the nonlinearity of inversion and the crosstalk between parameters are greatly reduced. In the third stage, the multistep-length gradient scheme is adopted. The density structure can be improved owing to the use of individual step lengths for different parameters. In the fourth stage, we make minor adjustments to the recovered P- and S-wave velocities and density by implementing conventional FWI again. Synthetic examples have determined that our hierarchical FWI scheme with the aforementioned steps obtains more plausible models than the conventional method. Inversion results of each stage and any three stages reveal that wavefield decomposition and the multistep-length approach are helpful to improve the accuracy of velocities and density, respectively, and all the stages of our hierarchical FWI method are necessary to give a good recovery of P- and S-wave velocities and density.

Advanced public transport and intelligent transport systems: new modelling challenges

ABSTRACTTransit system ‘big data’ collecting and processing, and bidirectional communication between transit travellers and information centres are emerging as two factors that enhance the tools supporting short-term forecasting of network status for transit operations control and for traveller information. However, the current methodologies applied in these tools do not seem to have reached the level of research in the field of transit network modelling. Therefore, several methodological issues connected to the development of such tools are analysed in this paper. These issues concern application and development of real-time on-board load short-term forecasting methods, real-time best path advice, real-time transit assignment modelling, individual path choice modelling, and real-time updating and upgrading of demand and supply model parameters.

Maximum Margin Dirichlet Process Mixtures for Clustering

The Dirichlet process mixtures (DPM) can automatically infer the model complexity from data. Hence it has attracted significant attention recently, and is widely used for model selection and clustering. As a generative model, it generally requires prior base distribution to learn component parameters by maximizing posterior probability. In contrast, discriminative classifiers model the conditional probability directly, and have yielded better results than generative classifiers.In this paper, we propose a maximum margin Dirichlet process mixture for clustering, which is different from the traditional DPM for parameter modeling. Our model takes a discriminative clustering approach, by maximizing a conditional likelihood to estimate parameters. In particular, we take a EM-like algorithm by leveraging Gibbs sampling algorithm for inference, which in turn can be perfectly embedded in the online maximum margin learning procedure to update model parameters. We test our model and show comparative results over the traditional DPM and other nonparametric clustering approaches.

Sensor Placement via Optimal Experiment Design in EMI Sensing of Metallic Objects

This work, under the optimal experimental design framework, investigates the sensor placement problem that aims to guide electromagnetic induction (EMI) sensing of multiple objects. We use the linearized model covariance matrix as a measure of estimation error to present a sequential experimental design (SED) technique. The technique recursively minimizes data misfit to update model parameters and maximizes an information gain function for a future survey relative to previous surveys. The fundamental process of the SED seeks to increase weighted sensitivities to targets when placing sensors. The synthetic and field experiments demonstrate that SED can be used to guide the sensing process for an effective interrogation. It also can serve as a theoretic basis to improve empirical survey operation. We further study the sensitivity of the SED to the number of objects within the sensing range. The tests suggest that an appropriately overrepresented model about expected anomalies might be a feasible choice.

Hybrid Wind Speed Prediction Based on a Self-Adaptive ARIMAX Model with an Exogenous WRF Simulation

Wind speed forecasting is difficult not only because of the influence of atmospheric dynamics but also for the impossibility of providing an accurate prediction with traditional statistical forecasting models that work by discovering an inner relationship within historical records. This paper develops a self-adaptive (SA) auto-regressive integrated moving average with exogenous variables (ARIMAX) model that is optimized very-short-term by the chaotic particle swarm optimization (CPSO) algorithm, known as the SA-ARIMA-CPSO approach, for wind speed prediction. The ARIMAX model chooses the wind speed result from the Weather Research and Forecasting (WRF) simulation as an exogenous input variable. Further, an SA strategy is applied to the ARIMAX process. When new information is available, the model process can be updated adaptively with parameters optimized by the CPSO algorithm. The proposed SA-ARIMA-CPSO approach enables the forecasting process to update training information and model parameters intelligently and adaptively. As tested using the 15-min wind speed data collected from a wind farm in Northern China, the improved method has the best performance compared with several other models.

Real-time Adaptive Kinematic Model Estimation of Concentric Tube Robots.

Kinematic models of concentric tube robots have matured from considering only tube bending to considering tube twisting as well as external loading. While these models have been demonstrated to approximate actual behavior, modeling error can be significant for medical applications that often call for positioning accuracy of 1-2mm. As an alternative to moving to more complex models, this paper proposes using sensing to adaptively update model parameters during robot operation. Advantages of this method are that the model is constantly tuning itself to provide high accuracy in the region of the workspace where it is currently operating. It also adapts automatically to changes in robot shape and compliance associated with the insertion and removal of tools through its lumen. As an initial exploration of this approach, a recursive on-line estimator is proposed and evaluated experimentally.

Component Model Parameter Updating for Landing Gear Linkages with Flexible Joints

This paper presents a practical and effective framework for updating landing gear finite element component models such that the updated reduced-order component mode synthesis model of the entire assembled system can provide accurate structural dynamic results in any possible landing gear configuration. By constructing component mode synthesis matrices using updated finite element parameters, the requirement to experimentally derive the modes required for component mode synthesis is avoided. Because it is impossible to directly measure joint frequency response functions, the experimental substructure boundary conditions are made to mimic natural joint-link connectivity by using flexible-hinged boundaries with links and joint hinges in arbitrary orientations. This exploits the fact that experimental frequency response functions obtained away from the joints can contain information about the joint dynamics, which result in resonance frequency shifts. The updating of each component model in multiple possible configurations is performed using a genetic algorithm with bounded inequality constraints on the updating parameters and nonlinear constraints on frequency response function parameters (natural frequencies, antiresonances, and fixture motion). Afterward, the updated flexible-hinged supports are analytically removed for final system component mode synthesis assembly. The effectiveness of the proposed framework is demonstrated with an experimental case study on a simplified landing gear model. Overall, the updating method can enable further dynamic testing of the complete system in a virtual environment, thereby reducing the need for extensive experimental testing on entire landing gear assemblies.

An online expectation maximization algorithm for exploring general structure in massive networks

Mixture model and stochastic block model (SBM) for structure discovery employ a broad and flexible definition of vertex classes such that they are able to explore a wide variety of structure. Compared to the existing algorithms based on the SBM (their time complexities are O(mc2), where m and c are the number of edges and clusters), the algorithms of mixture model are capable of dealing with networks with a large number of communities more efficiently due to their O(mc) time complexity. However, the algorithms of mixture model using expectation maximization (EM) technique are still too slow to deal with real million-node networks, since they compute hidden variables on the entire network in each iteration. In this paper, an online variational EM algorithm is designed to improve the efficiency of the EM algorithms. In each iteration, our online algorithm samples a node and estimates its cluster memberships only by its adjacency links, and model parameters are then estimated by the memberships of the sampled node and old model parameters obtained in the previous iteration. The provided online algorithm updates model parameters subsequently by the links of a new sampled node and explores the general structure of massive and growing networks with millions of nodes and hundreds of clusters in hours. Compared to the relevant algorithms on synthetic and real networks, the proposed online algorithm costs less with little or no degradation of accuracy. Results illustrate that the presented algorithm offers a good trade-off between precision and efficiency.

Self-modifying CA model using dual ensemble Kalman filter for simulating urban land-use changes

There are many different methods to calibrate cellular automata (CA) models for better simulation results of urban land-use changes. However, few studies have been reported on combination of parameter update and error control using local data in CA calibration procedures. This paper presents a self-modifying CA model (SM-CA) that uses the dual ensemble Kalman filter (dual EnKF), which enables the CA model to simultaneously update model parameters and simulation results by merging observation data (local data). We applied the proposed model to simulate urban land-use changes in a 13-year period (1993–2005) in Dongguan City, a rapidly urbanizing region in south China. Simulation results indicate that this model yields better simulation results than the conventional logistic-regression CA and decision-tree CA models. For example, the validation is carried out using cross-tabulation matrix. The simulation results of SM-CA have allocation disagreement of 10.18%, 19.64%, and 30.03% in 1997, 2001, and 2005, respectively, which are 2.12%, 2.47%, and 6% lower than conventional logistic-regression CA models.

Full waveform inversion in the frequency domain using direct iterative T-matrix methods

We present two direct iterative solutions to the nonlinear seismic waveform inversion problem that are based on volume integral equation methods for seismic forward modelling in the acoustic approximation. The solutions are presented in the frequency domain, where accurate inversion results can often be obtained using a relatively low number of frequency components. Our inverse scattering approach effectively replaces an ill-posed nonlinear inverse problem with a series of linear ill-posed inverse problems, for which there already exist efficient (regularized) solution methods. Both these solutions update the wavefield within the scattering domain after each iteration. The main difference is that the background medium Green functions are kept fixed in the first solution, but updated after each iteration in the second solution. This means that our solutions are very similar to the Born iterative (BI) and the distorted Born iterative (DBI) methods that are commonly used in acoustic and electromagnetic inverse scattering. However, we have eliminated the need to perform a full forward simulation (or to invert a huge matrix) at each iteration via the use of an iterative T-matrix method for fixed background media for the BI method and a variational T-matrix method for dynamic background media for the DBI method. The T-matrix (variation) is linearly related with the seismic wavefield data (residuals), but related with the unknown scattering potential model parameter (updates) in a non-linear manner, which is independent of the source-receiver configuration. This mathematical structure, which allows one to peel off the effects of the source-receiver configuration, is very attractive when dealing with multiple (simultaneous) sources, and is also compatible with the (future) use of renormalization methods for dealing with local minima problems. To illustrate the performance and potential of the two direct iterative methods for FWI, we performed a series of numerical experiments on synthetic seismic waveform data associated with a simple 2D model and the more complicated Marmousi model. The results of these numerical experiments suggest that the use of a fixed (e.g. smooth and ray-tracing friendly) background medium may be adequate for some applications with moderately large velocity contrasts, but the solution based on a dynamic (non-smooth and constantly updated) background medium will normally provide superiour inversion results; also in the case of low signal-to-noise ratios.

국내 태풍 예측

We develop a model to predict typhoons in Korea. We collect data for typhoons and classify those depending on the severity level. Following a Bayesian approach, we develop a model that explains the relationship between different levels of typhoons. Through the analysis of the model, we can predict the rate of typhoons, the probability of approaching Korean peninsular, and the probability of striking Korean peninsular. We show that the uncertainty for the occurrence of various types of typhoons reduces dramatically by adaptively updating model parameters as we acquire data.

Adaptive and on-line learning in non-stationary environments

It is important that the update of model parameters and structure be achieved without a “catastrophic forgetting”. Therefore, a balance between continuous learning and “forgetting” is necessary to deal with non-stationary environments. A considerable body of research has been previously devoted to the design of models (e.g. classifiers, cluster models, time-series models) whose operating environments are supposed to be static—there, the system behavior and operating conditions/modes in static environments does not change over time. However, this is a quite strong restriction as usually the characteristics of systems and environmental conditions change, evolve over time. A typical example of changing environments is the spam detection and filtering. The descriptions of the two classes “spam” and “nonspam” evolve over time due to the changes of user preferences and “spammers” techniques to trick spam classifiers. Another example is the behavior of human beings, which may be affected by different experiences, moods, daily conditions etc.—mimicking their cognitive capabilities in form of neural models thus require a permanent adaptation and regulation of the induced networks. In case of condition monitoring systems employing multi-sensor networks, new sensors may be added whose measured variables or classes should be ideally integrated on-the-fly into a large networks of identified models describing the relations and dependencies being present within the system. In many applications, the models rely on off-line training cycles from historic data samples, which are often costly to obtain or need to be pre-annotated to establish an initial model (e.g. in case of classifiers), especially when intending to guarantee a sufficient coverage of the feature space—issues which are cost-intensive for companies and thus decreasing the applicability and attraction of data-driven models in industrial and healthcare systems. Training the models on The computerization of many life activities and the advances in data collection and storage technology lead to obtain mountains of data. They are collected to capture information about a phenomena or a process behavior. These data are rarely of direct benefit. Thus, a set of techniques and tools are used to extract useful information for decision support, prediction, exploration and understanding of phenomena governing the data sources. Learning methods use historic data points about a process past behavior to build a predictor (classifier, regression model, time-series model). The latter is used as an old experience to predict the process future behavior. However, the predictor needs to adjust itself (self-correction or adaptation) as new events happen or new conditions/system states occur (e.g. during on-line operations). The goal is to ensure an accurate prediction of process behavior according to the changes in new incoming data characteristics. This requires a continuous learning over long period of time with the ability to evolve new structural components on demand and to forget data becoming obsolete and useless. Incremental and sequential learning are essential concepts in order to avoid time-intensive re-training phases and account for the systems dynamics/changing data characteristics with low computational effort and virtual memory usage (enhancing on-line performance). This is because data is processed in sample-wise and single-pass manner.

Application of the Multimodel Ensemble Kalman Filter Method in Groundwater System

With the development of in-situ monitoring techniques, the ensemble Kalman filter (EnKF) has become a popular data assimilation method due to its capability to jointly update model parameters and state variables in a sequential way, and to assess the uncertainty associated with estimation and prediction. To take the conceptual model uncertainty into account during the data assimilation process, a novel multimodel ensemble Kalman filter method has been proposed by incorporating the standard EnKF with Bayesian model averaging framework. In this paper, this method is applied to analyze the dataset obtained from the Hailiutu River Basin located in the northwest part of China. Multiple conceptual models are created by considering two important factors that control groundwater dynamics in semi-arid areas: the zonation pattern of the hydraulic conductivity field and the relationship between evapotranspiration and groundwater level. The results show that the posterior model weights of the postulated models can be dynamically adjusted according to the mismatch between the measurements and the ensemble predictions, and the multimodel ensemble estimation and the corresponding uncertainty can be quantified.

Parameterization of Offline and Online Hydraulic Simulation Models

The proper estimation and regular update of model parameters is crucial for the actuality of a mathematical simulation model representing the hydraulics and water quality of a real physical water distribution system. Especially, when the model is running online, uncertainties in model parameters can result in large discrepancies between model predictions and behavior of the real system. Therefore, adequate techniques for data acquisition, maintenance and update of model parameters have to be developed. In what follows the “parameterization” of a hydraulic online simulation model is described including the classification of parameters regarding data source, update cycles and function in the model. The parameterization framework was developed to benefit the modeling of water critical infrastructure systems.

Application of Artificial Intelligence for Bridge Deterioration Model.

The deterministic bridge deterioration model updating problem is well established in bridge management, while the traditional methods and approaches for this problem require manual intervention. An artificial-intelligence-based approach was presented to self-updated parameters of the bridge deterioration model in this paper. When new information and data are collected, a posterior distribution was constructed to describe the integrated result of historical information and the new gained information according to Bayesian theorem, which was used to update model parameters. This AI-based approach is applied to the case of updating parameters of bridge deterioration model, which is the data collected from bridges of 12 districts in Shanghai from 2004 to 2013, and the results showed that it is an accurate, effective, and satisfactory approach to deal with the problem of the parameter updating without manual intervention.

Goal-oriented updating of mechanical models

We introduce a goal-oriented procedure for the updating of mechanical models. It exploits as usual information coming from experimental data, but these are post-processed in a specific way in order to firstly update model parameters which are the most influent for the prediction of a given quantity of interest. The objective is thus to perform a partial model updating that enables to obtain an approximate value of the quantity of interest with sufficient accuracy and minimal model identification effort. The updating method uses the constitutive relation error framework, as well as duality and adjoint techniques, and defines dedicated cost functions. It leads to a convenient strategy that automatically selects the relevant parameter set to be updated. Performances of the approach are analyzed on examples involving linear elasticity and transient thermal models with possible noisy measurements.

Sequential Method with Incremental Analysis Update to Retrieve Leaf Area Index from Time Series MODIS Reflectance Data

High-quality leaf area index (LAI) products retrieved from satellite observations are urgently needed for crop growth monitoring and yield estimation, land-surface process simulation and global change studies. In recent years, sequential assimilation methods have been increasingly used to retrieve LAI from time series remote-sensing data. However, the inherent characteristics of these sequential assimilation methods result in temporal discontinuities in the retrieved LAI profiles. In this study, a sequential assimilation method with incremental analysis update (IAU) was developed to jointly update model states and parameters and to retrieve temporally continuous LAI profiles from time series Moderate Resolution Imaging Spectroradiometer (MODIS) reflectance data. Based on the existing multi-year Global Land Surface Satellite (GLASS) LAI product, a dynamic model was constructed to evolve LAI anomalies over time. The sequential assimilation method with an IAU technique takes advantage of the Kalman filter (KF) technique to update model parameters, uses the ensemble Kalman filter (EnKF) technique to update LAI anomalies recursively from time series MODIS reflectance data and then calculates the temporally continuous LAI values by combining the LAI climatology data. The method was tested over eight Committee on Earth Observing Satellites-Benchmark Land Multisite Analysis and Intercomparison of Products (CEOS-BELMANIP) sites with different vegetation types. The results indicate that the sequential method with IAU can precisely reconstruct the seasonal variation patterns of LAI and that the LAI profiles derived from the sequential method with IAU are smooth and continuous.

Goal-oriented updating of mechanical models using the adjoint framework

In this paper, we introduce a goal-oriented procedure for the updating of mechanical models. It is based as usual on information coming from measurement data, but these data are post-processed in a convenient way in order to firstly update model parameters which are the most influent for the prediction of a given quantity of interest. The objective is thus to perform a partial model calibration that enables to obtain an approximate value of the quantity of interest with sufficient accuracy and minimal model identification effort. The updating method uses the constitutive relation error framework as well as duality and adjoint techniques. It leads to a convenient strategy, mainly based on sensitivity analysis, that selects the relevant parameter set to be updated and also provides for useful quantitative tools in order to define optimal experiments. Performances of the approach are analyzed on examples involving linear elasticity and transient thermal models.

Updating of Travel Behavior Model Parameters and Estimation of Vehicle Trip Chain Based on Plate Scanning

This article proposes a maximum-likelihood method to update travel behavior model parameters and estimate vehicle trip chain based on plate scanning. The information from plate scanning consists of the vehicle passing time and sequence of scanned vehicles along a series of plate scanning locations (sensor locations installed on road network). The article adopts the hierarchical travel behavior decision model, in which the upper tier is an activity pattern generation model, and the lower tier is a destination and route choice model. The activity pattern is an individual profile of daily performed activities. To obtain reliable estimation results, the sensor location schemes for predicting trip chaining are proposed. The maximum-likelihood estimation problem based on plate scanning is formulated to update model parameters. This problem is solved by the expectation-maximization (EM) algorithm. The model and algorithm are then tested with simulated plate scanning data in a modified Sioux Falls network. The results illustrate the efficiency of the model and its potential for an application to large and complex network cases.