Dependent Normalized Random Measures

Changyou Chen, Vinayak Rao, Wray Buntine, Yee Whye Teh
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):969-977, 2013.

Abstract

In this paper we propose two constructions of dependent normalized random measures, a class of nonparametric priors over dependent probability measures. Our constructions, which we call mixed normalized random measures (MNRM) and thinned normalized random measures (TNRM), involve (respectively) weighting and thinning parts of a shared underlying Poisson process before combining them together. We show that both MNRM and TNRM are marginally normalized random measures, resulting in well understood theoretical properties. We develop marginal and slice samplers for both models, the latter necessary for inference in TNRM. In time-varying topic modelling experiments, both models exhibit superior performance over related dependent models such as the hierarchical Dirichlet process and the spatial normalized Gamma process.

Cite this Paper

BibTeX
@InProceedings{pmlr-v28-chen13i, title = {Dependent Normalized Random Measures}, author = {Chen, Changyou and Rao, Vinayak and Buntine, Wray and Whye Teh, Yee}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {969--977}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {3}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/chen13i.pdf}, url = {https://proceedings.mlr.press/v28/chen13i.html}, abstract = {In this paper we propose two constructions of dependent normalized random measures, a class of nonparametric priors over dependent probability measures. Our constructions, which we call mixed normalized random measures (MNRM) and thinned normalized random measures (TNRM), involve (respectively) weighting and thinning parts of a shared underlying Poisson process before combining them together. We show that both MNRM and TNRM are marginally normalized random measures, resulting in well understood theoretical properties. We develop marginal and slice samplers for both models, the latter necessary for inference in TNRM. In time-varying topic modelling experiments, both models exhibit superior performance over related dependent models such as the hierarchical Dirichlet process and the spatial normalized Gamma process.} }
Endnote
%0 Conference Paper %T Dependent Normalized Random Measures %A Changyou Chen %A Vinayak Rao %A Wray Buntine %A Yee Whye Teh %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-chen13i %I PMLR %P 969--977 %U https://proceedings.mlr.press/v28/chen13i.html %V 28 %N 3 %X In this paper we propose two constructions of dependent normalized random measures, a class of nonparametric priors over dependent probability measures. Our constructions, which we call mixed normalized random measures (MNRM) and thinned normalized random measures (TNRM), involve (respectively) weighting and thinning parts of a shared underlying Poisson process before combining them together. We show that both MNRM and TNRM are marginally normalized random measures, resulting in well understood theoretical properties. We develop marginal and slice samplers for both models, the latter necessary for inference in TNRM. In time-varying topic modelling experiments, both models exhibit superior performance over related dependent models such as the hierarchical Dirichlet process and the spatial normalized Gamma process.
RIS
TY - CPAPER TI - Dependent Normalized Random Measures AU - Changyou Chen AU - Vinayak Rao AU - Wray Buntine AU - Yee Whye Teh BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/05/26 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-chen13i PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 3 SP - 969 EP - 977 L1 - http://proceedings.mlr.press/v28/chen13i.pdf UR - https://proceedings.mlr.press/v28/chen13i.html AB - In this paper we propose two constructions of dependent normalized random measures, a class of nonparametric priors over dependent probability measures. Our constructions, which we call mixed normalized random measures (MNRM) and thinned normalized random measures (TNRM), involve (respectively) weighting and thinning parts of a shared underlying Poisson process before combining them together. We show that both MNRM and TNRM are marginally normalized random measures, resulting in well understood theoretical properties. We develop marginal and slice samplers for both models, the latter necessary for inference in TNRM. In time-varying topic modelling experiments, both models exhibit superior performance over related dependent models such as the hierarchical Dirichlet process and the spatial normalized Gamma process. ER -
APA
Chen, C., Rao, V., Buntine, W. & Whye Teh, Y.. (2013). Dependent Normalized Random Measures. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(3):969-977 Available from https://proceedings.mlr.press/v28/chen13i.html.

Related Material