Alexander Litvinenko, RWTH Aachen, Germany
THE SUBJECT SOFTWARE IS PROVIDED "AS IS" WITHOUT ANY WARRANTY
This project contains numerical examples for this article "Computing f-Divergences and Distances of High-Dimensional Probability Density Functions -- Low-Rank Tensor Approximations", available at https://arxiv.org/abs/2111.07164
The user needs, first, to install TT-Toolbox from here https://github.com/oseledets/TT-Toolbox .
Then to run TT-Toolbox/setup.m
Only after that the user can run single Matlabs files from this repository
The following functionality is available:
- Tensor train (TT) approximation of a probability characteristic function (pcf)
- TT approximation of a probability density function (pdf)
- Computation of the log() function of a pdf and then the KL distance
- Computation of the square root of a pdf and then the Hellinger distance
For example:
my_pdf_tensor_ex1.m was used for Example 6.1 and to generate data from Table 5, i.e.
for the computation of
my_H2dist_tensor.m was used to compute
my_KLD_tensor.m was used for computing
Other useful files:
KLD_simplified.m computes KLD between two \alpha stable distributions. Different ways are tested and compared with the analyticc solution.
Plese write me an email if you have any questions litvinenko@uq.rwth-aachen.de