A Latent Space Approach to Dynamic Embedding of Co-occurrence Data

A Latent Space Approach to Dynamic Embedding of Co-occurrence Data,Purnamrita Sarkar,Sajid M. Siddiqi,Geoffrey J. Gordon

A Latent Space Approach to Dynamic Embedding of Co-occurrence Data   (Citations: 9)
BibTex | RIS | RefWorks Download
We consider dynamic co-occurrence data, such as author-word links in papers pub- lished in successive years of the same con- ference. For static co-occurrence data, re- searchers often seek an embedding of the entities (authors and words) into a low- dimensional Euclidean space. We general- ize a recent static co-occurrence model, the CODE model of Globerson et al. (2004), to the dynamic setting: we seek coordinates for each entity at each time step. The coordi- nates can change with time to explain new observations, but since large changes are im- probable, we can exploit data at previous and subsequent steps to find a better explana- tion for current observations. To make in- ference tractable, we show how to approxi- mate our observation model with a Gaussian distribution, allowing the use of a Kalman filter for tractable inference. The result is the first algorithm for dynamic embedding of co-occurrence data which provides distri- butional information for its coordinate es- timates. We demonstrate our model both on synthetic data and on author-word data from the NIPS corpus, showing that it pro- duces intuitively reasonable embeddings. We also provide evidence for the usefulness of our model by its performance on an author- prediction task.
Published in 2007.
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...Given link data for T time periods, can we predict the relationships at time T +1? This problem has been considered in a variety of contexts [2], [3], [4]...
    • ...Modeling the time evolution of graphs has been considered, e.g., by Sakar et al. [4] who create time-evolving cooccurrence models that map entities into an evolving latent space...

    Evrim Acaret al. Link Prediction on Evolving Data Using Matrix and Tensor Factorization...

    • ...In the algorithm of Sarkar et al. [17], for example, word co-occurrence relations are analyzed instead of simple word frequency...

    Frank Van Hamet al. Mapping Text with Phrase Nets

Sort by: