Equivariance Properties in Machine Learning

Introduction When we think about the signs for what is a good model, we first look to notions of accuracy, precision, recall and the like. Evaluating these metrics typically require a ground truth to evaluate against, and is rarely the case that a machine learning practitioner will not have at least a few examples to test on. In medical imaging deep learning is applied to make measurements about our bodies such as organ size, which helps us gain understand our health.