Parameter Estimation Task Research Articles

Audio-Material Reconstruction for Virtualized Reality Using a Probabilistic Damping Model.

Modal sound synthesis has been used to create realistic sounds from rigid-body objects, but requires accurate real-world material parameters. These material parameters can be estimated from recorded sounds of an impacted object, but external factors can interfere with accurate parameter estimation. We present a novel technique for estimating the damping parameters of materials from recorded impact sounds that probabilistically models these external factors. We represent the combined effects of material damping, support damping, and sampling inaccuracies with a probabilistic generative model, then use maximum likelihood estimation to fit a damping model to recorded data. This technique greatly reduces the human effort needed and does not require the precise object geometry or the exact hit location. We validate the effectiveness of this technique with a comprehensive analysis of a synthetic dataset and a perceptual study on object identification. We also present a study establishing human performance on the same parameter estimation task for comparison.

Parameter estimation via weak measurement with machine learning

Weak measurement technique can provide high sensitivity for parameter estimation tasks. However in practice, time-varying errors induced by noises significantly compromise its advantages. While prolonging the measurement time can reduce the effect of short-term correlated noises (e.g. white noise), it will on the other hand increase the estimation errors in the presence of long-term correlated noises(e.g. 1/f and 1/f2 noises). The main obstacle of compensating these kind of errors is to precisely predict the time-varying trend of error, where the disturbances that causing error are highly unpredictable. In this work, we propose a weak measurement scheme with assistance of machine learning algorithm, which can ‘learn’ the time-varying trend of errors from data without an explicit model. In order to verify the feasibility of this scheme, we carry out a time-delay measurement experiment and achieve the noise compensation via machine learning algorithm. Experimental results demonstrate a 6 dB reduction of mean-square error comparing to the setup without machine learning, which suggest that our scheme can effectively improve the performance of weak measurement in the presence of long-time correlated noises.

Sparse Reconstruction Method for Separating Cardiac and Respiratory Components from Electrical Bioimpedance Measurements

In this work, we investigate the possibility of employing sparse reconstruction framework for the separation of cardiac and respiratory signal components from the bioimpedance measurements. The signal decomposition is complicated by the nonstationarity of the signal and overlapping of their spectra. The signal has a harmonic structure, which is sparse in the spectral domain. We approach the problem by considering a dictionary with integrated wideband elements describing spectral components of the considered signal. The parameter estimation task is solved by the means of sparse reconstruction where solving the optimization problem returns a sparse vector of relevant dictionary atoms.DOI: http://dx.doi.org/10.5755/j01.eie.24.5.21844

Sounderfeit: cloning a physical model using a conditional adversarial autoencoder


 
 
 An adversarial autoencoder conditioned on known parameters of a physical modeling bowed string syn- thesizer is evaluated for use in parameter estimation and resynthesis tasks. Latent dimensions are provided to cap- ture variance not explained by the conditional parameters. Results are compared with and without the adversarial training, and a system capable of “copying” a given parameter-signal bidirectional relationship is examined. A real- -time synthesis system built on a generative, conditioned and regularized neural network is presented, allowing to construct engaging sound synthesizers based purely on recorded data. 
 
 

Physics-aware Gaussian processes in remote sensing

Earth observation from satellite sensory data poses challenging problems, where machine learning is currently a key player. In recent years, Gaussian Process (GP) regression has excelled in biophysical parameter estimation tasks from airborne and satellite observations. GP regression is based on solid Bayesian statistics, and generally yields efficient and accurate parameter estimates. However, GPs are typically used for inverse modeling based on concurrent observations and in situ measurements only. Very often a forward model encoding the well-understood physical relations between the state vector and the radiance observations is available though and could be useful to improve predictions and understanding. In this work, we review three GP models that respect and learn the physics of the underlying processes in the context of both forward and inverse modeling. After reviewing the traditional application of GPs for parameter retrieval, we introduce a Joint GP (JGP) model that combines in situ measurements and simulated data in a single GP model. Then, we present a latent force model (LFM) for GP modeling that encodes ordinary differential equations to blend data-driven modeling and physical constraints of the system governing equations. The LFM performs multi-output regression, adapts to the signal characteristics, is able to cope with missing data in the time series, and provides explicit latent functions that allow system analysis and evaluation. Finally, we present an Automatic Gaussian Process Emulator (AGAPE) that approximates the forward physical model using concepts from Bayesian optimization and at the same time builds an optimally compact look-up-table for inversion. We give empirical evidence of the performance of these models through illustrative examples of vegetation monitoring and atmospheric modeling.

Secure Information Sharing in Adversarial Adaptive Diffusion Networks

In this paper, information sharing over adversarial adaptive networks is considered. Defense against adversarial attacks is enforced by an attack detection mechanism that guides the diffusion strategy in the parameter estimation task and the transmission decisions of agents. The latter are taken to maximize long-term rewards, expressed in terms of estimation performance, communication cost, and illegal gains in a nonzero sum repeated game with imperfect action monitoring. In order to track agents’ behavior and stimulate cooperation, a reputation protocol is designed, along with the deployment of a set of honest nodes whose behavior is consistent with the reputation protocol. An algorithm that promotes cooperation and can achieve secure parameter estimation is described and illustrated by simulations.

Investigation of nonlinear epidemiological models for analyzing and controlling the MERS outbreak in Korea

Much concern has arisen regarding serious epidemics due to the Middle East Respiratory Syndrome (MERS) coronavirus. The first MERS case of Korea was reported on 20 May 2015, and since then, the MERS outbreak in Korea has resulted in hundreds of confirmed cases and tens of deaths. Deadly infectious diseases such as MERS have significant direct and indirect social impacts, which include disease-induced mortality and economic losses. Also, a delayed response to the outbreak and underestimating its danger can further aggravate the situation. Hence, an analysis and establishing efficient strategies for preventing the propagation of MERS is a very important and urgent issue. In this paper, we propose a class of nonlinear susceptible-infectious-quarantined (SIQ) models for analyzing and controlling the MERS outbreak in Korea. For the SIQ based ordinary differential equation (ODE) model, we perform the task of parameter estimation, and apply optimal control theory to the controlled SIQ model, with the goal of minimizing the infectious compartment population and the cost of implementing the quarantine and isolation strategies. Simulation results show that the proposed SIQ model can explain the observed data for the confirmed cases and the quarantined cases in the MERS outbreak very well, and the number of the MERS cases can be controlled reasonably well via the optimal control approach.

Compressive parameter estimation via K-median clustering

Compressive sensing (CS) has attracted significant attention in parameter estimation tasks, where parametric dictionaries (PDs) collect signal observations for a sampling of the parameter space and yield sparse representations for signals of interest when the sampling is dense. While this sampling also leads to high dictionary coherence, one can leverage structured sparsity models to prevent highly coherent dictionary elements from appearing simultaneously in the recovered signal. However, the resulting approaches depend heavily on the careful setting of the maximum allowable coherence; furthermore, their guarantees are not concerned with general parameter estimation performance. We propose the use of earth mover’s distance (EMD), as applied to a pair of true and estimated PD coefficient vectors, to measure the parameter estimation error. We formally analyze the connection between the EMD and the parameter estimation error and show that the EMD provides a better-suited metric for parameter estimation performance than the Euclidean distance. Additionally, we analyze the previously described relationship between K-median clustering and EMD-optimal sparse approximation and leverage it to develop improved PD-based parameter estimation algorithms. Numerical experiments verify our theoretical results, showing that the proposed compressive parameter estimation algorithms have performance similar to state-of-the-art algorithms while featuring simpler implementation and broader applicability.

A Hybrid Optimization-Based Approach for Parameter Estimation and Investigation of Fractional Dynamics in Ultracapacitors

The increasing demand of electrical energy storage has motivated the wide usage of ultracapacitors. Ultracapacitors are capable of storing and delivering energy at a rate much higher than conventional rechargeable batteries. The dynamics characteristics of ultracapacitors are better exhibited by fractional calculus as compared to the conventional calculus. The present work aims at estimating the fractional model parameters of an ultracapacitor using experimental data and further investigating the dependency of fractional parameters on the operating conditions. The parameter estimation task has been formulated as an optimization problem which aims at minimizing the deviation between the model and experimental output using a hybrid optimization technique. The hybrid algorithm combines an improved version of seeker optimization algorithm for global exploration with a local search technique, i.e., Nelder---Mead simplex search. The results show that the hybrid algorithm is capable of identifying the model parameters efficiently in both time and frequency domain. The fractional behavior has been found to be dependent on the initial condition, magnitude and offset of the applied voltage.

Minimum Proper Loss Estimators for Parametric Models

In this paper, we propose a methodology for systematically deriving estimators minimizing proper loss functions defined on parametric statistical models, by restricting ourselves to losses taking a functional form which allows for straightforward proofs of key properties. Proving propriety is tantamount to deriving a pairwise divergence quantity between probability measures, admitting a natural interpretation of minimum proper loss estimators as divergence minimizers. We show that proper losses of varying complexity can be readily constructed given propriety of losses taking a rudimentary form, and that for many important models, verifying desired properties in the rudimentary case is immediate. As a special case, we derive computationally tractable estimators requiring only the first and second moments, verify strict propriety properties, and empirically confirm their utility though parameter estimation tasks using both controlled simulations and real-world meteorological network data sets. Comparisons against numerous standard estimators show that the proposed estimators are at least competitive with, and often markedly superior to all standard references, uniformly across tasks, data sets, and model classes, suggesting a strong alternative to standard benchmarks in a wide variety of estimation problems.

NARMAX model identification using a set-theoretic evolutionary approach

A method is presented for identification of system models that are linear in parametric structure, but arbitrarily nonlinear in signal operations. The fundamental parameter estimation task uses a set-theoretic analysis of the data to deduce feasible sets of solutions in light of certain model assumptions. In turn, measurable set solution properties are used to assess the viability of nonlinear regressor functions that “compete for survival” as components of the model best fit to represent the system. The strategy blends traditional system identification methods with three modeling strategies that are not commonly employed in signal processing: linear-time-invariant-in-parameters models, set-based parameter identification, and evolutionary selection of the model structure. The algorithm can identify nonlinear model structure and estimate parameters in the presence of different unknown noise scenarios, especially correlated noise. This paper reports the theoretical foundation of the methods and the simulation studies to illustrate the performance benefits.

Optimal Initial State for Fast Parameter Estimation in Nonlinear Dynamical Systems

This paper deals with optimal initial state for parameter estimation in a bounded error context. Based on sensitivity analysis, our original method uses this information to evaluate the optimal initial state start point for precise parameter estimation task. In our framework, the uncertainty (on measurement noise and parameters) is considered as to be interval, thus guaranteed sensitivity analysis method has been applied which ensure to obtain a better parameter estimation. Once the system's initial state is determined, a set inversion procedure combined with a volumetric criterion's contractor is proposed. The proposition of check the necessity of the contractor's action moment is new. Besides, we point out that the set membership computation should have the measurement points as few as possible, an elementary effect analysis in the interval analysis context is also implemented to achieve the fast and guaranteed parameter estimation aim.

Inference of reactive transport model parameters using a Bayesian multivariate approach

AbstractParameter estimation of subsurface transport models from multispecies data requires the definition of an objective function that includes different types of measurements. Common approaches are weighted least squares (WLS), where weights are specified a priori for each measurement, and weighted least squares with weight estimation (WLS(we)) where weights are estimated from the data together with the parameters. In this study, we formulate the parameter estimation task as a multivariate Bayesian inference problem. The WLS and WLS(we) methods are special cases in this framework, corresponding to specific prior assumptions about the residual covariance matrix. The Bayesian perspective allows for generalizations to cases where residual correlation is important and for efficient inference by analytically integrating out the variances (weights) and selected covariances from the joint posterior. Specifically, the WLS and WLS(we) methods are compared to a multivariate (MV) approach that accounts for specific residual correlations without the need for explicit estimation of the error parameters. When applied to inference of reactive transport model parameters from column‐scale data on dissolved species concentrations, the following results were obtained: (1) accounting for residual correlation between species provides more accurate parameter estimation for high residual correlation levels whereas its influence for predictive uncertainty is negligible, (2) integrating out the (co)variances leads to an efficient estimation of the full joint posterior with a reduced computational effort compared to the WLS(we) method, and (3) in the presence of model structural errors, none of the methods is able to identify the correct parameter values.

A Review on Parameter Estimation Techniques of Software Reliability Growth Models

Software reliability is considered as a quantifiable metric, which is defined as the probability of a software to operate without failure for a specified period of time in a specific environment. Various software reliability growth models have been proposed to predict the reliability of a software. These models help vendors to predict the behaviour of the software before shipment. The reliability is predicted by estimating the parameters of the software reliability growth models. But the model parameters are generally in nonlinear relationships which creates many problems in finding the optimal parameters using traditional techniques like Maximum Likelihood and least Square Estimation. Various stochastic search algorithms have been introduced which have made the task of parameter estimation, more reliable and computationally easier. Parameter estimation of NHPP based reliability models, using MLE and using an evolutionary search algorithm called Particle Swarm Optimization, has been explored in the paper.

A Hybrid Parameter Estimation Algorithm for S-System Model of Gene Regulatory Networks

The reconstruction of a gene regulatory network expressed in terms of a S-system model may be accomplished by a simple task of parameter estimation. Empirical data indicate that biological gene networks are sparsely connected and the average number of upstream-regulators per gene is less than two, implying that most of parameter variables in the S-system model are zero. It is thus desired to search for a parameter estimation algorithm that is capable of identifying the connectivity of the gene network and determining its reduced number of non-zero parameters. A hybrid algorithm is presented for identification and parameter estimation of gene network structure described by a S-system model. It combines an optimization process with a system identification method commonly used in the aerospace community. Constraint equations in a matrix form are formulated to deal with the steady state and the network connectivity conditions. The system parameter vector resides in the null space of the constraint matrix. The resulting network structure and system parameters are optimally tuned by minimizing the error of state time history. A numerical experiment is given to illustrate the hybrid parameter estimation algorithm.

Statistically-based approach for monitoring of micro-seismic events

We consider the problem of optimal statistical estimation of micro-seismic source parameters using multichannel data from surface arrays of seismometers affected by a strong seismic noise. The problem is treated as a statistical task of parameter estimation for a general type multidimensional linear model with random or completely unknown input time functions. The maximum-likelihood generic estimators are derived and their relationship with the well-known seismic emission tomography (SET) algorithm is established. The proposed estimation algorithms perform processing of multichannel discrete observations in the frequency domain and can be implemented in on-line mode. Using the method of successive independent trials (Monte-Carlo), we demonstrate that a proposed statistical estimation algorithm provides much higher accuracy of the micro-seismic event source location in real noise conditions than the widely used SET algorithm and, as a consequence, is more reliable in practical applications.

Linear and nonlinear model updating of reinforced concrete T-beam bridges using artificial neural networks

The key parameters affecting dynamic and static responses of structural systems often change during their life cycles due to aging, deterioration, damage and rehabilitation. Model updating is a major research field that investigates numerical methods to improve simulation ability of finite element (FE) models by identifying the modified parameters in structural systems based on data collected from field experiments and/or laboratory tests. In this paper, artificial neural networks (ANNs) are used to develop an efficient method for finite element (FE) model updating of reinforced concrete (RC) T-beam bridges. The FE model of a sample bridge selected from Pennsylvania’s bridge population is calibrated using neural networks trained according to datasets generated from linear and non-linear analyses separately. The simulated responses obtained from calibrated FE models are compared to the field-measured responses of the bridge to quantify accuracy of parameter estimation and success of the model updating process. The present study evinces the fact that ANNs can still be used efficiently and reliably for parameter estimation tasks under a high level of uncertainty and complexity that arises from aging and deterioration of RC bridges as well as nonlinear material properties of concrete. The study also indicates significance of non-linear response analysis for parameter identification for RC bridges, and underlines that only consideration of dynamic responses for model updating may lead to erroneous parameter predictions especially when the calibration is based on linear bridge responses.

Automatic termination of parallel optimization runs of global stochastic optimization methods by upper limit criterion

Many combinations of adjustable parameters should be tested in optimization experiments of biochemical networks to find the smallest subset of parameters enabling the best improvements of objective function both in case of design task and parameter estimation task. In case of optimization with global stochastic optimization methods one of the problems is the termination of the optimization run looking for a good compromise between spent computational resources and probability that the best found value of objective function will be the global optimum. Longer runs increase the possibility to each the global optimum. Automatic termination criteria in case of consensus or stagnation of parallel optimization runs have been proposed as criteria for automatic termination. Varying the consensus and delay time settings different probability of reaching global optimum and duration of optimization can be reached. It is proposed to modify automatic optimization termination criteria of parallel optimization runs applying upper limit agreement of a number of parallel optimization runs. Automatic application of upper limit agreement would reduce the duration of scanning of the whole space of combination of adjustable parameters. This approach is tested on the yeast glycolysis model with six adjustable parameters using COPASI, CoRunner and ConvAn software for five parallel optimization runs per combination of adjustable parameters.

Identification of regulatory structure and kinetic parameters of biochemical networks via mixed-integer dynamic optimization

BackgroundRecovering the network topology and associated kinetic parameter values from time-series data are central topics in systems biology. Nevertheless, methods that simultaneously do both are few and lack generality.ResultsHere, we present a rigorous approach for simultaneously estimating the parameters and regulatory topology of biochemical networks from time-series data. The parameter estimation task is formulated as a mixed-integer dynamic optimization problem with: (i) binary variables, used to model the existence of regulatory interactions and kinetic effects of metabolites in the network processes; and (ii) continuous variables, denoting metabolites concentrations and kinetic parameters values. The approach simultaneously optimizes the Akaike criterion, which captures the trade-off between complexity (measured by the number of parameters), and accuracy of the fitting. This simultaneous optimization mitigates a possible overfitting that could result from addition of spurious regulatory interactions.ConclusionThe capabilities of our approach were tested in one benchmark problem. Our algorithm is able to identify a set of plausible network topologies with their associated parameters.

Determining Model Accuracy Requirements for Automotive Engine Coldstart Hydrocarbon Emissions Control

In this work, a systematic method is introduced to determine the required accuracy of an automotive engine model used for real-time optimal control of coldstart hydrocarbon (HC) emissions. The engine model structure and development are briefly explained and the model predictions versus experimental results are presented. The control design problem is represented with a dynamic optimization formulation on the basis of the engine model and solved using the Pontryagin’s minimum principle (PMP). To relate the level of plant/model mismatch and the control performance degradation in practice, a sensitivity analysis using a computationally efficient method is employed. In this way, the sensitivities or the effects of small parameter variations on the optimal solution, which is the minimum of cumulative tailpipe HC emissions over the coldstart period, are calculated. There is a good agreement between the sensitivity analysis results and the experimental data. The sensitivities indicate the directions of the subsequent parameter estimation and model improvement tasks to enhance the control-relevant accuracy, and thus, the control performance. Furthermore, they provide some insights to simplify the engine model, which is critical for real-time implementation of the coldstart optimal control system.