Combining model outputs into a combined model output

Abstract

The invention relates to a prediction system () for applying multiple trained models to an input instance, for example, for detection or segmentation of objects in a medical image. The multiple trained models form a combined model. A trained model determines a model output for an input instance by determining a representation of the input instance in a common latent space and determining the respective model output therefrom. The combined model further comprises dataset fingerprints of the multiple trained models, characterizing representations of its training instances in the latent space. To determine an output of the combined model for an input instance, correspondence scores are determined between the input instance and the multiple trained models, indicating correspondences between the input instance and the respective training dataset in the latent space. The combined model output is determined by combining respective models according to the correspondence scores.