Inductive Venn-Abers and related regressors
TLDR
This paper generalizes Venn-Abers predictors to unbounded regression using conformal prediction, showing improved efficiency for larger datasets.
Key contributions
- Extends Venn-Abers predictors from binary classification/bounded regression to unbounded regression.
- Incorporates conformal prediction to enable Venn-Abers for unbounded regression tasks.
- Empirically demonstrates improved predictive efficiency of derived point regressors.
- Shows performance gains over standard regressors, especially with larger training datasets.
Why it matters
This work extends Venn-Abers predictors to unbounded regression, addressing a key limitation. It offers a new method for valid probabilistic predictions in a broader range of tasks. The improved predictive efficiency for larger datasets is a significant practical benefit.
Original Abstract
Venn-Abers predictors are probabilistic predictors that enjoy appealing properties of validity, but their major limitation is that they are applicable only to the case of binary classification, with a recent extension to bounded regression. We generalize them to the case of unbounded regression, which requires adding an element of conformal prediction. In our simulation and empirical studies we investigate the predictive efficiency of point regressors derived from Venn-Abers regressors and argue that they somewhat improve the predictive efficiency of standard regressors for larger training sets.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.