TY - GEN
T1 - An integrated model for bayesian learning of sparse representation and classifier training
AU - Li, Jin
AU - Hua, Gang
AU - Lan, Xuguang
AU - Zheng, Nanning
PY - 2013
Y1 - 2013
N2 - We present an integrated model for Bayesian learning of sparse representation and classifier training, and apply it for the task of visual recognition. Most previous work learns the sparse representation and trains the classifier on top of it in two separate steps. We cast these two into a unified probabilistic model. This way, the supervised labels can effectively affect the learning of the sparse representation. In the training phase, the inference of the joint expectation for dictionary, code, classifier and other variables under the observation of descriptors and labels is carried out by Gibbs Sampling. In the testing phase, based on the learned parameters, the sparse code and the class label of the image are obtained by Bayesian inference. The proposed model is evaluated on Caltech 101 dataset and its efficacy is demonstrated by a careful analysis of the experimental results.
AB - We present an integrated model for Bayesian learning of sparse representation and classifier training, and apply it for the task of visual recognition. Most previous work learns the sparse representation and trains the classifier on top of it in two separate steps. We cast these two into a unified probabilistic model. This way, the supervised labels can effectively affect the learning of the sparse representation. In the training phase, the inference of the joint expectation for dictionary, code, classifier and other variables under the observation of descriptors and labels is carried out by Gibbs Sampling. In the testing phase, based on the learned parameters, the sparse code and the class label of the image are obtained by Bayesian inference. The proposed model is evaluated on Caltech 101 dataset and its efficacy is demonstrated by a careful analysis of the experimental results.
UR - https://www.scopus.com/pages/publications/84894155869
U2 - 10.1007/978-3-319-03731-8_57
DO - 10.1007/978-3-319-03731-8_57
M3 - 会议稿件
AN - SCOPUS:84894155869
SN - 9783319037301
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 620
EP - 628
BT - Advances in Multimedia Information Processing, PCM 2013 - 14th Pacific-Rim Conference on Multimedia, Proceedings
PB - Springer Verlag
T2 - 14th Pacific-Rim Conference on Multimedia, PCM 2013
Y2 - 13 December 2013 through 16 December 2013
ER -