Data Scaling Laws in NMT: The Effect of Noise and Architecture.
Yamini BansalBehrooz GhorbaniAnkush GargBiao ZhangMaxim KrikunColin CherryBehnam NeyshaburOrhan FiratPublished in: CoRR (2022)
Keyphrases
- data sets
- data analysis
- noisy data
- experimental data
- high quality
- data collection
- input data
- real time
- data distribution
- missing data
- synthetic data
- computer systems
- data processing
- data sources
- data structure
- training data
- data quality
- statistical analysis
- signal to noise ratio
- database
- data points
- raw data
- spatial data
- random noise
- historical data
- noise reduction
- data objects
- original data
- data acquisition
- databases
- decision trees
- attribute values
- database systems
- image restoration
- prior knowledge
- end users
- edge detection
- knowledge discovery