Bayesian Nonparametric Model for Power Disaggregation
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Zoubin Ghahramani.
We propose an infinite factorial unboundedstate hidden Markov model (HMM) through the construction of a Bayesian nonparametric (BNP) prior over integervalued matrices (in which each column represents a Markov chain) with the property of presenting an infinite number of columns with an unbounded number of states, namely, IFUHMM . First, we extend the existent infinite factorial binarystate HMM to allow for any number of states, and derive two Markov chain Monte Carlo (MCMC) and a variational inference algorithms. Then, we modify this model to allow an unbounded number of states and derive a new inference algorithm based on MCMC methods that properly deals with the tradeoff between the unbounded number of states and chains. Finally, we apply both the infinite factorial (nonbinary) HMM and the IFUHMM to the power disaggregation problem and show their ability to provide more interpretable representations of the data structure than the existing BNP approaches for HMMs.
This talk is part of the Machine Learning @ CUED series.
This talk is included in these lists:
Note that exdirectory lists are not shown.
