We introduce the Randomized Dependence Coefficient (RDC), a novel measure of non-linear dependence between random variables of arbitrary dimension. Throughout theoretical analysis and empirical evaluation, it is shown that RDC is a highly-scalable (running over millions of samples in seconds), very easy to implement (five lines of R code) estimator of the well-known Hirschfeld-Gebelein-R\'enyi's Maximum Correlation Coefficient.
View on arXiv