MACEst

What is MACEst? MACEst stands for Model Agnostic Confidence Estimator. It is an algorithm that can estimate confidence in the predictions made by machine learning models. The algorithm uses a set of nearest neighbours and is different from other methods in that it calculates confidence as a local quantity that takes into account both aleatoric and epistemic uncertainty. This is different from standard calibration methods, which use a global point prediction model as a starting point for the con

1 / 1