Leveraging Topology and Intrinsic Dimension for Effective Regression Representation Learning

Effective regression representation learning requires the feature space to be topologically similar to the target space and have the same intrinsic dimension as the target space.