this is for holding javascript data
Xavier Holt edited Introduction_and_contribution_1_page__.md
almost 8 years ago
Commit id: d1677f4e968766b8f2384e401ca704bdbbc3c0be
deletions | additions
diff --git a/Introduction_and_contribution_1_page__.md b/Introduction_and_contribution_1_page__.md
index 73c3491..7e3252d 100644
--- a/Introduction_and_contribution_1_page__.md
+++ b/Introduction_and_contribution_1_page__.md
...
## # Introduction, and contribution (1 page)
## Timeline Model
We adopt a nonparametric Bayesian approach to the TLG problem. As argued above, this is useful for a number of reasons. Having an underlying topic structure is a useful abstraction in and of itself. It simplifies queries such as related-stories and story-importance. The hierarchal nature of the model also allows us to choose our desired level of granularity. We largely adopt the formulation of Ahmed et al.\cite{Ahmed2011}. We do so because it is a general framework with demonstrated application to several domains. It also operates in a streaming manner, and is therefore particularly appropriate to our task.
## Contributions
### Contributions to the Model
Our inference method is one of our primary contributions to the state of the art. Inference in nonparametric Bayesian formulations can largely be divided into sampling and expectation-maximisation (EM) like approaches. The former has been applied to TLG but as of yet no attempt has been made to apply the latter. The work of Wang et al. \cite{Wang2011} and Bryant et al.\cite{Bryant2012} on variational inference is a step in this direction. They develop a variational framework for the hierarchal dirichlet model, a fundamental part of all nonparametric TLG formulations. As such we seek to build on their work and apply it to specifically the TLG case. Our goal is motivated by the excellent performance of variational inference. This is both generally \cite{Grimmer_2010} and specifically; Wang et al.\cite{Wang2011} had excellent performance on a dataset of 400,000 articles, an order of magnitude larger than any sampling-based inference on the TLG problem.
\cite{Minard:2015tw}
