One For All & All For One: Bypassing Hyperparameter Tuning with Model Averaging for Cross-Lingual Transfer.
Fabian David SchmidtIvan VulicGoran GlavasPublished in: EMNLP (Findings) (2023)
Keyphrases
- cross lingual
- model averaging
- transfer learning
- machine translation
- bayesian methods
- posterior distribution
- gaussian processes
- hyperparameters
- language modeling
- text classification
- supervised classification
- model selection
- cross validation
- maximum a posteriori
- news articles
- closed form
- bayesian framework
- machine learning
- gaussian process
- document clustering
- language model
- parameter settings
- generative model
- co occurrence
- knowledge discovery
- probability distribution