We introduced a novel data-driven neighborhood sampling approach, learned by a Reinforcement Learning, replacing random sampling with uniform distribution in GraphSAGE (Hamilton et al. (2017)). In order to embed nodes in a large-scale graph using li…
We show that BERT is a Markov random field language model. We give a practical algorithm for generating from BERT without any additional training and verify in experiments that the algorithm produces diverse and fairly fluent generations. Further wo…
We have described a simple adaptation of BERT as a passage re-ranker that has become the state of the art on two different tasks, which are TREC-CAR and MS MARCO. We have made the code to reproduce our MS MARCO entry publicly available.
In our attempt to replicate the grammar induction results reported in Shen et al. (2018), we find several experimental design problems that make the results difficult to interpret. However, in experiments and analyses going well beyond the scope of …
Credentials & highlights
- Project Views