About

All contents from https://youngjookim.github.io/sdr/ were submitted to and accepted by the Information Visualization journal.

High-dimensional Sharpened Dimensionality Reduction (HD-SDR)

High-dimensional Sharpened Dimensionality Reduction (HD-SDR) sharpens the original high-dimensional data before dimensionality reduction to create visually segregrated sample clusters, which is targeted for user-guided labeling.

Setup

Code is written in C++ and is available here.

https://github.com/youngjookim/sdr

To get the code running please include the following:
a. Nanoflann header file
b. Eigen 3.3.3
c. Tapkee dimension reduction library

The sharpening filter is implemented in function ‘sharpening_for_dr()’ in main.cpp.

Input files should be located in the ‘Data’ folder and consist of one header file and one data file for each data set. Please run ‘example_data_set_1’.

Experimental results

All the experimental results for this project are found here.

Supplemental materials

Supplemental materials are found here.