Success of Uncertainty-Aware Deep Models Depends on Data Manifold Geometry.
Mark PenrodHarrison TermottoVarshini ReddyJiayu YaoFinale Doshi-VelezWeiwei PanPublished in: CoRR (2022)
Keyphrases
- data analysis
- historical data
- data sets
- original data
- high quality
- synthetic data
- data processing
- raw data
- experimental data
- uncertain data
- accurate models
- prior knowledge
- data structure
- database
- models built
- computer systems
- learned models
- statistical analysis
- data sources
- training data
- statistical models
- neural network
- statistical methods
- physical models
- learning models
- missing values
- data distribution
- missing data
- high dimensional data
- labeled data
- model selection
- input data
- probabilistic model
- decision trees