Probabilistic Classification

Main.ProbabilisticClassification History

Hide minor edits - Show changes to output

May 15, 2017, at 05:50 PM by Vovk - added an example of a forecasting system
Changed lines 1-3 from:
In ''[[https://en.wikipedia.org/wiki/Probabilistic_classification | probabilistic classification]]'' problems, the label space $\mathbf{Y}$ is finite and [[Forecaster]] is required to output a probability measure on $\mathbf{Y}$.  In ''binary classification problems'', the size of $\mathbf{Y}$ is 2; usually, $\mathbf{Y}=\{0,1\}$ or $\mathbf{Y}=\{-1,1\}$, and a forecast can be interpreted as a number in the interval ${[0,1]}$ (the predicted probability of the outcome 1).
to:
In ''[[https://en.wikipedia.org/wiki/Probabilistic_classification | probabilistic classification]]'' problems, the label space $\mathbf{Y}$ is finite and [[Forecaster]] is required to output a probability measure on $\mathbf{Y}$.  In ''binary classification problems'', the size of $\mathbf{Y}$ is 2; usually, $\mathbf{Y}=\{0,1\}$ or $\mathbf{Y}=\{-1,1\}$, and a forecast can be interpreted as a number in the interval ${[0,1]}$ (the predicted probability of the outcome 1).

An example of forecasting strategies for probabilistic classification is various [[Venn predictor]]s
.
May 15, 2017, at 05:44 PM by Vovk - created the page
Added line 1:
In ''[[https://en.wikipedia.org/wiki/Probabilistic_classification | probabilistic classification]]'' problems, the label space $\mathbf{Y}$ is finite and [[Forecaster]] is required to output a probability measure on $\mathbf{Y}$.  In ''binary classification problems'', the size of $\mathbf{Y}$ is 2; usually, $\mathbf{Y}=\{0,1\}$ or $\mathbf{Y}=\{-1,1\}$, and a forecast can be interpreted as a number in the interval ${[0,1]}$ (the predicted probability of the outcome 1).