Filters








181 Hits in 4.9 sec

Identification of Probability weighted ARX models with arbitrary domains [article]

Alessandro Brusaferri and Matteo Matteucci and Stefano Spinelli
2020 arXiv   pre-print
Then, the parameters of both the ARX submodels and the classifier are concurrently estimated by maximizing the likelihood of the overall model using Expectation Maximization.  ...  To this end, we propose a method based on a probabilistic mixture model, where the discrete state is represented through a multinomial distribution conditioned by the input regressors.  ...  The Adam algorithm starts with a learning rate of 0.01, running for 3 epochs mini-batches of 100 samples in each M-iteration. The network starts from random normal weights with zero-mean.  ... 
arXiv:2009.13975v1 fatcat:ktlu4amaajgnbbdmyjodipd5vi

Hierarchical Bayesian ARX models for robust inference

Johan Dahlin, Fredrik Lindsten, Thomas B. Schön, Adrian Wills
2012 IFAC Proceedings Volumes  
We consider these models in a Bayesian setting and perform inference using numerical procedures based on Markov Chain Monte Carlo methods.  ...  We demonstrate that this choice of distribution for the innovations provides an increased robustness to data anomalies, such as outliers and missing observations.  ...  Similar models are also considered by Christmas and Everson [2011] , who derive a variational Bayes algorithm for the inference problem.  ... 
doi:10.3182/20120711-3-be-2027.00318 fatcat:rgy2zkjw6ragrodpzvdjo3bhbi

Linear parameter varying model identification with an unknown scheduling variable in presence of missing observations- a robust global approach

Xianqiang Yang, Xin Liu, Boxuan Han
2018 IET Control Theory & Applications  
This study focuses on identifying the linear parameter varying (LPV) system with an unknown scheduling variable in the presence of missing measurements and the system output data contaminated with outliers  ...  The formulations of the proposed algorithm are finally derived in the expectation-maximisation algorithm scheme and the formulas to estimate the unknown parameters of LPV ARX model and scheduling variable  ...  Yang and Yin [15] proposed a variational Bayesian inference method for identification of finite impulse response models with randomly missing measurements and the distributions, not just the single point  ... 
doi:10.1049/iet-cta.2017.1176 fatcat:35j4baslxnbkzd3f5tsrpljqyq

Aeroelastic validation and Bayesian updating of a downwind wind turbine

Rakesh Sarma, Richard P. Dwight, Axelle Viré
2020 Wind Energy  
Aerodynamic forces from prescribed forced-motion simulations are used to train a time-domain autoregressive with exogenous input (ARX) model with a localised forcing term, which provides accurate and cheap  ...  The industrial design procedures mostly employ simpler models (of lower fidelity than RANS such as vortex lattice method for aircraft design), which results in missing physical effects, This is an open  ...  This allows the trained ARX model based on the baseline model to be used for estimating the aerodynamic forces in the probabilistic framework.  ... 
doi:10.1002/we.2448 fatcat:coxw4vwxq5djzaizyueyjdt67q

Identification of Time-Varying Autoregressive Systems Using Maximum a Posteriori Estimation

Tesheng Hsiao
2008 IEEE Transactions on Signal Processing  
Based on these probabilistic constraints, an iterative algorithm is proposed to evaluate the maximum a posteriori estimates of the parameters.  ...  It is formulated as a Bayesian inference problem with constraints on the conditional and prior probabilities of the time-varying parameters.  ...  Gemez and Maravall explored the state space representation of the autoregressive-integrated-moving-average (ARIMA) models with missing observations.  ... 
doi:10.1109/tsp.2008.919393 fatcat:nc6h26vw6bgfjlojmcv2ekfxuu

Distributive Model-Based Sensor Fault Diagnosis in Wireless Sensor Networks

Chun Lo, Mingyan Liu, Jerome P. Lynch
2013 2013 IEEE International Conference on Distributed Computing in Sensor Systems  
With the Yule-Walker equation based fast iterative ARX coefficient training method, the ARX training process can be done in a reasonable time by a 8-bit micro-controller (Atmel ATmega128) running at 8MHz  ...  For the training of the ARX pair-wise time series models, all three excitation types are utilized with a unique ARX model found for each excitation and measurement type.  ... 
doi:10.1109/dcoss.2013.58 dblp:conf/dcoss/LoLL13 fatcat:q4kx7l2kiba6djvfcyinz5rjm4

An Over-Sampling Amplitude-Limited Variational Bayesian Method for the Identification of Hammerstein Model

Baochang Xu, Likun Yuan, Yaxin Wang
2020 IEEE Access  
, and autoregressive exogenous (ARX) model with random missing output data [30] .  ...  An Amplitude-Limited Variational Bayesian (ALVB) method combined with the over-sampling closedloop structure with colored noise for multivariable nonlinear models is proposed.  ...  .: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS model which is applicable for colored noise.  ... 
doi:10.1109/access.2020.3044272 fatcat:pc4jtogzi5g3vg2qka5grtndzi

A state-space approach to sparse dynamic network reconstruction [article]

Zuogong Yue, Johan Thunberg, Lennart Ljung, Jorge Goncalves
2018 arXiv   pre-print
., using ARX models) requires a large amount of parameters in model selection.  ...  To solve the SBL problem, another EM algorithm is embedded, where we impose conditions on network identifiability in each iteration.  ...  CONCLUSIONS AND OUTLOOK This paper proposes an algorithm to reconstruct sparse dynamic networks using the EM algorithm embedded with sparse Bayesian learning.  ... 
arXiv:1811.08677v1 fatcat:3d27dyzu7rafxdh7e4izhtmi6m

Electricity price forecasting in European Day Ahead Markets: a greedy consideration of market integration

Ties Van der Heijden, Jesus Lago, Peter Palensky, Edo Abraham
2021 IEEE Access  
We apply the algorithm to build price forecasting models for the Dutch market, using candidate countries selected through an integrated analysis based on open-source European electricity market data.  ...  Two types of models (LEAR and the Deep Neural Network) are considered for the DAM price forecasting with and without European features.  ...  for every iteration and per model type.  ... 
doi:10.1109/access.2021.3108629 fatcat:3a7i5ib5orcbdnx2fkeagt7cva

PARAMETRIC TIME-DOMAIN METHODS FOR THE IDENTIFICATION OF VIBRATING STRUCTURES—A CRITICAL COMPARISON AND ASSESSMENT

K.A. PETSOUNIS, S.D. FASSOIS
2001 Mechanical systems and signal processing  
Concise summaries of the methods, highlighting their principles and realisations, are provided, while the study is based upon a six-degree-of-freedom structural model characterised by two closely spaced  ...  A critical comparison of four stochastic (PEM, 2SLS, LMS, IV) and three deterministic (LS, Prony, ERA) methods for the parametric time-domain identi"cation of vibrating structures from random excitation  ...  The BIC exhibits limited variation for k*34, although its minimum is attained for k"64. An ARX(34, 34) model is "nally selected.  ... 
doi:10.1006/mssp.2001.1424 fatcat:vlbcwtqsvfgjzmrt73grs64x5q

Use of Kullback-Leibler divergence for forgetting

Miroslav Kárný, Josef Andrýsek
2009 International Journal of Adaptive Control and Signal Processing  
Approximations based on the KLD with the reversed order of arguments preserves this property.  ...  This practical result is of importance for adaptive systems and opens a way for improving the functional approximation.  ...  INTRODUCTION Recursive estimation of finitely parameterized models of random input-output relationships is the core of adaptive systems. Bayesian methodology [1, 2] solves it consistently.  ... 
doi:10.1002/acs.1080 fatcat:2hswrgus7bfeniwguls3mfyqjy

Sparse Bayesian Inference of Multivariable ARX Networks [article]

J. Jin, Y. Yuan, A. Webb, J. Goncalves
2019 arXiv   pre-print
The paper considers a class of sparse linear time-invariant networks where the dynamics are represented by multivariable ARX models.  ...  GESBL combines Sparse Bayesian Learning (SBL) and Group Sparse Bayesian Learning (GSBL) to introduce penalties for complexity, both in terms of element (system order of nonzero connections) and group sparsity  ...  To avoid this limitation, we resort to Sparse Bayesian Learning. Remark 2: In practice, most systems are better modelled with ARMAX models instead of just ARX.  ... 
arXiv:1605.09543v2 fatcat:7gtdmo34orfbtlkoc6pit3wtym

Time Series Data Cleaning with Regular and Irregular Time Intervals [article]

Xi Wang, Chen Wang
2020 arXiv   pre-print
At present, to deal with these time series containing errors, besides keeping original erroneous data, discarding erroneous data and manually checking erroneous data, we can also use the cleaning algorithm  ...  Data with errors could not be stored in the database, which results in the loss of data assets. Handling the dirty data in time series is non-trivial, when given irregular time intervals.  ...  In fact, Bayesian prediction model is a technique based on Bayesian statistics.  ... 
arXiv:2004.08284v3 fatcat:ze4c5bjhvjhdvdlykgn6c6mil4

MIMO LMS-ARMAX IDENTIFICATION OF VIBRATING STRUCTURES—PART I: THE METHOD

S.D. FASSOIS
2001 Mechanical systems and signal processing  
A critical assessment of the LMS-ARMAX method under various noise conditions, as well as comparisons with a simpler ARX version and the ERA (Eigensystem Realization Algorithm), are undertaken based upon  ...  A comprehensive linear multi stage autoregressive moving average with exogenous excitation (LMS-ARMAX) method for e!  ...  D (B) O H H D ( j) ) BH"C\(B) ) B(B) [s;m]. (9) The algorithm starts by estimating an s-variate ARX representation of truncated order p of the above form, that is, y[t]# N H H W ( j) ) y[t!  ... 
doi:10.1006/mssp.2000.1382 fatcat:ao3czvqluzgilaivkdbeo7sev4

On evolutionary system identification with applications to nonlinear benchmarks

K. Worden, R.J. Barthorpe, E.J. Cross, N. Dervilis, G.R. Holmes, G. Manson, T.J. Rogers
2018 Mechanical systems and signal processing  
which combine the insight of any prior physical-law based models (white box) with the power of machine learners with universal approximation properties (black box).  ...  As one might expect from the 'no-free-lunch' theorem for optimisation, the methodology is not particularly sensitive to the particular (EO) algorithm used, and a number of different variants are presented  ...  In addition, Tim Rogers is grateful to Ramboll Oil and Gas for financial support and Geoff Holmes similarly thanks Innovate UK.  ... 
doi:10.1016/j.ymssp.2018.04.001 fatcat:y3rrs753v5ddzodkewnd54bnea
« Previous Showing results 1 — 15 out of 181 results