Conformal Prediction in Learning Under Privileged Information Paradigm with Applications in Drug Discovery

TL;DR

This paper investigates conformal prediction within the learning under privileged information paradigm, demonstrating its validity and efficiency benefits in certain datasets, including drug discovery, by leveraging SVM+ models for reliable prediction intervals.

Contribution

It introduces the integration of conformal prediction with SVM+ in the LUPI framework and evaluates its effectiveness on benchmark and drug discovery datasets.

Findings

Privileged information improves model validity and efficiency.

SVM+ enables valid prediction intervals at specified significance levels.

Improvement varies across datasets, with limited gains in drug discovery applications.

Abstract

This paper explores conformal prediction in the learning under privileged information (LUPI) paradigm. We use the SVM+ realization of LUPI in an inductive conformal predictor, and apply it to the MNIST benchmark dataset and three datasets in drug discovery. The results show that using privileged information produces valid models and improves efficiency compared to standard SVM, however the improvement varies between the tested datasets and is not substantial in the drug discovery applications. More importantly, using SVM+ in a conformal prediction framework enables valid prediction intervals at specified significance levels.