Filters








8,314 Hits in 4.5 sec

Discriminative training of GMM-HMM acoustic model by RPCL learning

Zaihu Pang, Shikui Tu, Dan Su, Xihong Wu, Lei Xu
2011 Frontiers of Electrical and Electronic Engineering in China  
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  ...  This approach is featured by embedding a rival penalized competitive learning (RPCL) mechanism on the level of hidden Markov states.  ...  [10] for training GMM components across different hidden Markov states.  ... 
doi:10.1007/s11460-011-0152-0 fatcat:gu57rzxrnbg27la3eaqlghjm5i

Learning a Hidden Markov Model-Based Hyper-heuristic [chapter]

Willem Van Onsem, Bart Demoen, Patrick De Causmaecker
2015 Lecture Notes in Computer Science  
A new approach to hyper-heuristics is proposed that addresses this problem by modeling and learning hyper-heuristics by means of a hidden Markov Model.  ...  A simple model shows how a reasonable update scheme for the probability vector by which a hyper-heuristic chooses the next heuristic leads to neglecting useful mutation heuristics.  ...  Acknowledgements This research is funded by the Institute for Innovation through Science and Technology (IWT) under grant 131 751.  ... 
doi:10.1007/978-3-319-19084-6_7 fatcat:h5fotiu2orbghjs7a5vw7legqy

A Markov Model-Based Fusion Algorithm for Distorted Electronic Technology Archives

Lei Wang, Jun Ye
2022 Computational Intelligence and Neuroscience  
The feasibility and accuracy of the online algorithm for the parameter estimation problem of the model are illustrated through numerical experiments with two specific examples of hidden Markov models,  ...  This paper presents an in-depth study and analysis of the restoration of distorted electronic technology archives using Markov models and proposes a corresponding fusion algorithm.  ...  into three basic forms: the continuous hidden Markov model, the discrete hidden Markov model, and the semicontinuous hidden Markov model. e three basic problems described above are all for discrete hidden  ... 
doi:10.1155/2022/4202181 pmid:35498173 pmcid:PMC9054407 fatcat:tbxu3e63n5g57cmk5f72iklipm

Accurate statistical spoken language understanding from limited development resources

Ivan V. Meza-Ruiz, Sebastian Riedel, Oliver Lemon
2008 Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing  
For example, by representing such features in MLNs, but without using a gazetteer, we outperform the Hidden Vector State (HVS) model of He and Young 2006 (1.26% improvement, a 13% error reduction).  ...  We show that it is possible to reach state-of-the-art performance using minimal additional resources, by using Markov Logic Networks (MLNs).  ...  For n > 0 we have a linear chain model which uses an n th order Markov assumption.  ... 
doi:10.1109/icassp.2008.4518786 dblp:conf/icassp/Meza-RuizRL08 fatcat:kq3h2mszgzgrzmsqr7t5acqgs4

Decoding As Dynamic Programming For Recurrent Autoregressive Models

Najam Zaidi, Trevor Cohn, Gholamreza Haffari
2020 International Conference on Learning Representations  
Decoding in autoregressive models (ARMs) consists of searching for a high scoring output sequence under the trained model.  ...  The MAP solution is then used to recreate an improved factor graph approximation of the ARM via updated auxiliary variables.  ...  The authors are grateful to the anonymous reviewers for their insightful comments and corrections.  ... 
dblp:conf/iclr/ZaidiCH20 fatcat:xbpikdazffdtpmovuxgqkckz5u

Detecting Unsafe Driving Patterns using Discriminative Learning

Yue Zhou, Wei Xu, Huazhong Ning, Yihong Gong, Thomas S. Huang
2007 Multimedia and Expo, 2007 IEEE International Conference on  
The fusion is performed using a discriminatively trained graphical model -conditional random field (CRF). The proposed approach offers several advantage over existing information fusing approaches.  ...  We propose a discriminative learning approach for fusing multichannel sequential data with application to detect unsafe driving patterns from multi-channel driving recording data.  ...  A computational state-based model of driver behavior is developed in [5] using Hidden Markov Model (HMM) and after training it is able to predict when the driver is about to brake or turn.  ... 
doi:10.1109/icme.2007.4284929 dblp:conf/icmcs/ZhouXNGH07 fatcat:li7vfrz6jrah5ar3bw6fdfa6yi

A Distribution Similarity Based Regularizer for Learning Bayesian Networks [article]

Weirui Kong, Wenyi Wang
2018 arXiv   pre-print
With proper restrictions, learned models usually generalize better.  ...  For many problems, common information exists among those factors. Adding similarity restrictions can be viewed as imposing prior knowledge for model regularization.  ...  Multi-task Learning Multi-Task Learning (MTL) is a learning paradigm in machine learning and its target is to leverage useful information contained in multiple related tasks to help improve the generalization  ... 
arXiv:1808.06347v1 fatcat:lygjldo4h5gxxji752l3ggq5zq

Human-computer dialogue simulation using hidden Markov models

H. Cuayahuitl, S. Renals, O. Lemon, H. Shimodaira
2005 IEEE Workshop on Automatic Speech Recognition and Understanding, 2005.  
Our method uses a network of Hidden Markov Models (HMMs) to predict system and user intentions, where a "language model" predicts sequences of goals and the component HMMs predict sequences of intentions  ...  This paper presents a probabilistic method to simulate task-oriented human-computer dialogues at the intention level, that may be used to improve or to evaluate the performance of spoken dialogue systems  ...  Therefore, we use Input Hidden Markov Models (IHMMs) and Input-Output Hidden Markov Models (IOHMMs), which are extensions of the standard HMMs [14] , see figures 2c and 2d.  ... 
doi:10.1109/asru.2005.1566485 fatcat:4hxmph4nrjav7ksbpzhootvqx4

Information extraction from research papers using conditional random fields

Fuchun Peng, Andrew McCallum
2006 Information Processing & Management  
This paper makes an empirical exploration of several factors, including variations on Gaussian, exponential and hyperbolic-L 1 priors for improved regularization, and several classes of features and Markov  ...  With the increasing use of research paper search engines, such as CiteSeer, for both literature search and hiring decisions, the accuracy of such systems is of paramount importance.  ...  Empirical Study Hidden Markov Models Here we also briefly describe a HMM model we used in our experiments.  ... 
doi:10.1016/j.ipm.2005.09.002 fatcat:hjmczooa2jgbxhp36pu2ttzuqm

The Impact of Task-Oriented Feature Sets on HMMs for Dialogue Modeling

Kristy Elizabeth Boyer, Eunyoung Ha, Robert Phillips, James C. Lester
2011 SIGDIAL Conferences  
Human dialogue serves as a valuable model for learning the behavior of dialogue systems.  ...  Hidden Markov models' sequential structure is well suited to modeling human dialogue, and their theoretical underpinnings are consistent with the conception of dialogue as a stochastic process with a layer  ...  These sequences correspond to the input observations for learning an HMM. Hidden Markov Models HMMs explicitly model hidden states within a doubly stochastic structure (Rabiner, 1989) .  ... 
dblp:conf/sigdial/BoyerHPL11 fatcat:jybaflxoxzaw3obpkdrttorikq

Markov random fields for textures recognition with local invariant regions and their geometric relationships

J. Blanchet, F. B. P. Forbes, C. Schmid
2005 Procedings of the British Machine Vision Conference 2005  
We show that recognition can be improved by using a Hidden Markov Model (HMM) as organizational model when learning the texture classes.  ...  We showed that Hidden Markov Models were natural candidates and focused on a texture recognition task for illustration.  ... 
doi:10.5244/c.19.72 dblp:conf/bmvc/BlanchetFS05 fatcat:2u4g2t3o6zbvjf6mdhfrk45fkm

Comparison of Large Margin Training to Other Discriminative Methods for Phonetic Recognition by Hidden Markov Models

Fei Sha, Lawrence K. Saul
2007 2007 IEEE International Conference on Acoustics, Speech and Signal Processing - ICASSP '07  
In this paper we compare three frameworks for discriminative training of continuous-density hidden Markov models (CD-HMMs).  ...  Unlike CML and MCE, our formulation of large margin training explicitly penalizes incorrect decodings by an amount proportional to the number of mislabeled hidden states.  ...  INTRODUCTION Most modern speech recognizers are built from continuous-density hidden Markov models (CD-HMMs).  ... 
doi:10.1109/icassp.2007.366912 dblp:conf/icassp/ShaS07 fatcat:tz6ns4orxrd6fcgflwtjp2bm3e

Minimum classification error learning for sequential data in the wavelet domain

D. Tomassi, D.H. Milone, L. Forzani
2010 Pattern Recognition  
In these models, hidden Markov trees account for local dynamics in a multiresolution framework, while standard hidden Markov models capture longer correlations in time.  ...  Numerical experiments on phoneme recognition show important improvement over the recognition rate achieved by the same models trained using maximum likelihood estimation.  ...  A left-to-right hidden Markov model uses hidden Markov trees as models for the observed data in the wavelet domain. 15 sinc(i) Research Center for Signals, Systems and Computational Intelligence (fich.unl.edu.ar  ... 
doi:10.1016/j.patcog.2010.07.010 fatcat:m3sy4qyfgfekdkrpvlm5hfkk4m

Incorporating non-local information into information extraction systems by Gibbs sampling

Jenny Rose Finkel, Trond Grenager, Christopher Manning
2005 Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics - ACL '05  
Most current statistical natural language processing models use only local features so as to permit dynamic programming in inference, but this makes them unable to fully account for the long distance structure  ...  We show how to solve this dilemma with Gibbs sampling, a simple Monte Carlo method used to perform approximate inference in factored probabilistic models.  ...  Additionally, we would like to that our reviewers for their helpful comments.  ... 
doi:10.3115/1219840.1219885 dblp:conf/acl/FinkelGM05 fatcat:cktkw2haxzbz5klcv2elfr4d3i

Extended Probabilistic Latent Semantic Analysis for Automatic Image Annotation

Dongping Tian
2017 Journal of Information Hiding and Multimedia Signal Processing  
On the other hand, since the traditional expectation-maximization (EM) algorithm used to train the PLSA model is sensitive to its initialization, so a rival penalized competitive learning (RPCL) based  ...  On one hand, the traditional bag-of-visual-words model is improved by integrating the contextual semantic information among visual words based on the PLSA model.  ...  The author would like to sincerely thank the anonymous reviewers for their valuable comments and insightful suggestions that have helped to improve the paper.  ... 
dblp:journals/jihmsp/Tian17a fatcat:3sgydjovbvbgfdulhkj725o3eu
« Previous Showing results 1 — 15 out of 8,314 results