Radar HRRP statistical recognition with temporal factor analysis by automatic Bayesian Ying-Yang harmony learning

Radar high-resolution range profiles(HRRPs)are typical high-dimensional and interdimension dependently distributed data,the statistical modeling of which is a challenging task for HRRP-based target recognition.Supposing that HRRP samples are independent and jointly Gaussian distributed,a recent work[Du L,Liu H W,Bao Z.IEEE Transactions on Signal Processing,2008,56(5):1931–1944]applied factor analysis(FA)to model HRRP data with a two-phase approach for model selection,which achieved satisfactory recognition performance.The theoretical analysis and experimental results reveal that there exists high temporal correlation among adjacent HRRPs.This paper is thus motivated to model the spatial and temporal structure of HRRP data simultaneously by employing temporal factor analysis(TFA)model.For a limited size of high-dimensional HRRP data,the two-phase approach for parameter learning and model selection suffers from intensive computation burden and deteriorated evaluation.To tackle these problems,this work adopts the Bayesian Ying-Yang(BYY)harmony learning that has automatic model selection ability during parameter learning.Experimental results show stepwise improved recognition and rejection performances from the twophase learning based FA,to the two-phase learning based TFA and to the BYY harmony learning based TFA with automatic model selection.In addition,adding many extra free parameters to the classic FA model and thus becoming even worse in identifiability,the model of a general linear dynamical system is even inferior to the classic FA model.

Codimensional matrix pairing perspective of BYY harmony learning:hierarchy of bilinear systems,joint decomposition of data-covariance,and applications of network biology

One paper in a preceding issue of this journal has introduced the Bayesian Ying-Yang(BYY)harmony learning from a perspective of problem solving,parameter learning,and model selection.In a complementary role,the paper provides further insights from another perspective that a co-dimensional matrix pair(shortly co-dim matrix pair)forms a building unit and a hierarchy of such building units sets up the BYY system.The BYY harmony learning is re-examined via exploring the nature of a co-dim matrix pair,which leads to improved learning performance with refined model selection criteria and a modified mechanism that coordinates automatic model selection and sparse learning.Besides updating typical algorithms of factor analysis(FA),binary FA(BFA),binary matrix factorization(BMF),and nonnegative matrix factorization(NMF)to share such a mechanism,we are also led to(a)a new parametrization that embeds a de-noise nature to Gaussian mixture and local FA(LFA);(b)an alternative formulation of graph Laplacian based linear manifold learning;(c)a codecomposition of data and covariance for learning regularization and data integration;and(d)a co-dim matrix pair based generalization of temporal FA and state space model.Moreover,with help of a co-dim matrix pair in Hadamard product,we are led to a semi-supervised formation for regression analysis and a semi-blind learning formation for temporal FA and state space model.Furthermore,we address that these advances provide with new tools for network biology studies,including learning transcriptional regulatory,Protein-Protein Interaction network alignment,and network integration.

Discriminative training of GMM-HMM acoustic model by RPCL learning

This paper presents a new discriminative approach for training Gaussian mixture models(GMMs)of hidden Markov models(HMMs)based acoustic model in a large vocabulary continuous speech recognition(LVCSR)system.This approach is featured by embedding a rival penalized competitive learning(RPCL)mechanism on the level of hidden Markov states.For every input,the correct identity state,called winner and obtained by the Viterbi force alignment,is enhanced to describe this input while its most competitive rival is penalized by de-learning,which makes GMMs-based states become more discriminative.Without the extensive computing burden required by typical discriminative learning methods for one-pass recognition of the training set,the new approach saves computing costs considerably.Experiments show that the proposed method has a good convergence with better performances than the classical maximum likelihood estimation(MLE)based method.Comparing with two conventional discriminative methods,the proposed method demonstrates improved generalization ability,especially when the test set is not well matched with the training set.

Learning Gaussian mixture with automatic model selection:A comparative study on three Bayesian related approaches

Three Bayesian related approaches,namely,variational Bayesian(VB),minimum message length(MML)and Bayesian Ying-Yang(BYY)harmony learning,have been applied to automatically determining an appropriate number of components during learning Gaussian mixture model(GMM).This paper aims to provide a comparative investigation on these approaches with not only a Jeffreys prior but also a conjugate Dirichlet-Normal-Wishart(DNW)prior on GMM.In addition to adopting the existing algorithms either directly or with some modifications,the algorithm for VB with Jeffreys prior and the algorithm for BYY with DNW prior are developed in this paper to fill the missing gap.The performances of automatic model selection are evaluated through extensive experiments,with several empirical findings:1)Considering priors merely on the mixing weights,each of three approaches makes biased mistakes,while considering priors on all the parameters of GMM makes each approach reduce its bias and also improve its performance.2)As Jeffreys prior is replaced by the DNW prior,all the three approaches improve their performances.Moreover,Jeffreys prior makes MML slightly better than VB,while the DNW prior makes VB better than MML.3)As the hyperparameters of DNW prior are further optimized by each of its own learning principle,BYY improves its performances while VB and MML deteriorate their performances when there are too many free hyper-parameters.Actually,VB and MML lack a good guide for optimizing the hyper-parameters of DNW prior.4)BYY considerably outperforms both VB and MML for any type of priors and whether hyper-parameters are optimized.Being different from VB and MML that rely on appropriate priors to perform model selection,BYY does not highly depend on the type of priors.It has model selection ability even without priors and performs already very well with Jeffreys prior,and incrementally improves as Jeffreys prior is replaced by the DNW prior.Finally,all algorithms are applied on the Berkeley segmentation database of real world images.Again,BYY considerably outperforms both VB and MML,especially in detecting the objects of interest from a confusing background.