توسعه ي ميدان عصبي شرطي بر اساس تركيب خبرگان براي برچسب زدن داده هاي داراي توالي

چكيده انگليسي :

Extending of Conditional Neural Field based on Mixture of Experts for Labeling of Sequence Data Tahereh Mahdipour t mahdipour@ec iut ac ir May 29 2017 Department of Electrical and Computer Engineering Isfahan Unuversity of Technology Isfahan 84156 83111 Iran Degree M Sc Language Persian Supervisors Dr Mehran Safayani safayani@cc iut ac irAbstract Sequence labeling is one of the important problems in pattern recognition which involves assignment of labels toeach member of various kinds of sequences like sequences of characters images or speech One approach for sequencelabeling is to model input output structure as a graphical model Farther analysis can be performed on the designedgraphical model Graphical models are powerful tools for modeling probability distributions with large number ofvariables In some problems like handwriting recognition the relation between input and output features can be nonlinearand highly complicated Initially generative models such as Hidden Markov Model HMM were commonly used forsequence labeling After a while another model called Conditional random field CRF was represented which becamepopular quickly because of its capabilities in resolving some issues related to previous generative models CRF is adiscriminative probabilistic model Experiments in recent years show that combining CRF with other models increasesthe performance In this thesis the combination of the Conditional Random Field model and the concept of mixture of experts isinvestigated A mixture of experts model increases the learning accuracy through partitioning the input space and havinga focused expert network for every partition It has been shown that utilizing mixture of experts model in learning amodel will increase its performance In this research by using a number of expert networks which are some typesof neural networks between the input and output layers of a CRF model a higher level of features is obtained fromthe observation sequences and used for training the model A clustering algorithm is used to assign input strings toexperts To do this due to the inequality of the length of the observation sequences clustering is initially performedon the elements of each sequence After this there will be two choices for assigning experts to input data In the firstchoice by voting among the clusters of the elements of a string its cluster is determined and the entire elements ofthat sequence will be used for learning the related expert and model parameters In the second choice according to thecluster related to each element the training can be performed on all experts assigned to clusters of the input elements The result of these two choices gives us two models Experimental results in the application of handwriting recognition demonstrate that the proposed models can con siderably improve recognition accuracy in comparison to previous models In this research the comparison is performedwith models such as neural networks conditional random field and conditional neural field The results indicate thatthe first and second proposed models improve the recognition accuracy up to 7 and 7 5 respectively Keywords Sequence Labeling Log Linear Model Discriminative Model ConditionalRandom Field Mixture of Experts

مهدي پور، طاهره

توسعه ي ميدان عصبي شرطي بر اساس تركيب خبرگان براي برچسب زدن داده هاي داراي توالي

برچسب زني رشته , مدل , Log-Linear مدل افتراقي , ميدان تصادفي شرطي , تركيب خبرگان