Generalised Isotonic Regression

This page provides quick access to an Emscripten compiled version of the C++ library here. For anything more than a few thousand points, you should prefer the library or contained CLI. With this tool, however, you can visualise how the space is partitioned with each subsequent iteration of the algorithm. For more background on the algorithm see the paper Generalized Isotonic Regression or my attempt at explaining it here.

  1. Paste your data into the "Input" box. Examples of the supported format can be seen by clicking on one of the "Example" buttons.
  2. Click the "Run Regression" button to run the algorithm with the currently selected loss function. The result at each iteration can be displayed by selecting a chosen iteration from the drop-down box above the "Result".
  3. The "Plot" button can then be used to visualise how the space is partitioned at each iteration. This only supports 2- and 3-dimensional data.