Ensemble

State-of-the-art and high-performance medical image classification pipelines are heavily utilizing Ensemble Learning strategies.

The idea of Ensemble Learning is to assemble diverse models or multiple predictions and thus boost prediction performance.

AUCMEDI currently supports the following Ensemble Learning techniques:

Technique	Description
Augmenting	Inference Augmenting (test-time augmentation) function for augmenting unknown images for prediction.
Bagging	Cross-Validation based Bagging for equal models trained with different sampling.
Stacking	Ensemble of unequal models with a fitted Metalearner stacked on top of it.
Composite	Combination of Stacking and Bagging via cross-validation with a fitted Metalearner stacked on top of it.

Info

More information on performance impact of Ensemble Learning in medical image classification can be found here:

Dominik Müller, Iñaki Soto-Rey, Frank Kramer. (2022) An Analysis on Ensemble Learning optimized Medical Image Classification with Deep Convolutional Neural Networks. arXiv e-print: https://arxiv.org/abs/2201.11440