ensembles
Multiple models either with varying parameters or sees varied training data. Collectively, these sub-models learn parts of the problem and can be combined to give better accuracy as well as quantify #uncertainty.
The ensemble framework can be either [[bagging]] or [[boosting]].
Backlinks
Shapley values
4. Additivty—Shapley values are additive, such that they can be used in [[ensembles]].
scalable-uncertainties-from-deep-ensembles
- Interpretation of [[dropout]] uncertainties by Gal and Ghahramani as both [[MCMC]] sampling, as well as "creation" of [[ensembles]] of neural networks.