Inductive Venn Prediction

Venn Predictors (VPs) are machine learning algorithms that can provide well calibrated multiprobability outputs for their predictions. An important drawback of Venn Predictors is their computational inefficiency, especially in the case of large datasets. In this work, we investigate and propose Inductive Venn Predictors (IVPs), which can overcome the computational inefficiency problem of the original Transductive Venn Prediction framework. We develop an IVP algorithm and perform a detailed comparison of its time efficiency, accuracy, and quality of probabilistic outputs with those of the original Transductive Venn Predictor (TVP). The results demonstrate that our method provides well calibrated results while maintaining high accuracy. The IVP outperforms the original TVP method in terms of time efficiency, while also providing well-calibrated probabilistic estimates. Another observation is that the probability intervals of the IVP are tighter than those of the TVP.