RDC: Randomized Dependence Coefficient

The RDC is a measure of nonlinear dependence between two (possibly multidimensional) variables.

Algorithm by: David Lopez-Paz, Philipp Hennig, and Bernhard Schoelkopf (2013 paper).

Code by: Pascal Pfeil, inspired by Gary Doran's python rdc implementation.

Usage

#include "rdc/rdc.h"

// ...

const std::vector<double> x_values{/* ... */};
const std::vector<double> y_values{/* ... */};

rdc::rdc(x_values, y_values) 

x and y are currently only supported to be 1-D vectors of any type that supports sorting with std::sort. They have to be of the same size.

There are additional keyword parameters for rdc that correspond to parameters described in the paper.

CMake

You can include rdc in your own CMake based project like this:

include(FetchContent)
set(RDC_BUILD_EXAMPLE OFF)
set(RDC_BUILD_TESTS OFF)
set(RDC_BUILD_BENCHMARKS OFF)
FetchContent_Declare(
    rdc
    GIT_REPOSITORY "https://github.com/pascalpfeil/rdc.git"
    GIT_TAG main
)
FetchContent_MakeAvailable(rdc)
FetchContent_GetProperties(rdc SOURCE_DIR RDC_INCLUDE_DIR)
include_directories(${RDC_INCLUDE_DIR})

target_link_libraries(your_target rdc)

Header

The rdc/rdc.h header has just one dependency, Eigen. If you don't use CMake, you can also provide Eigen yourself and copy the header into your project.