Multi-armed bandits for resource efficient, online optimization of language model pre-training: the use case of dynamic masking.
Iñigo UrteagaMoulay-Zaïdane DraïdiaTomer LancewickiShahram KhadiviPublished in: ACL (Findings) (2023)
Keyphrases
- expectation maximization
- language model
- probabilistic model
- mixture model
- language modeling
- multi armed bandits
- n gram
- context sensitive
- online learning
- document retrieval
- speech recognition
- retrieval model
- language modelling
- statistical language models
- query expansion
- ad hoc information retrieval
- smoothing methods
- relevance model
- dirichlet prior
- bayesian networks
- test collection
- training set
- information retrieval
- pseudo relevance feedback
- query terms
- model selection
- supervised learning
- upper bound
- language model for information retrieval