cca-vs-sammon-opposite-distance-focus

IN premise — summaries/2026/08/24/wiki-Nonlinear_dimensionality_reduction-chunk-2.md

Created 2026-08-24T17:11:20+00:00

Curvilinear Component Analysis (CCA) focuses on preserving small output-space distances (its stress function relates to a sum of right Bregman divergences), while Sammon's mapping focuses on preserving small input-space distances

Summary

CCA and Sammon's mapping have opposite priorities about which distances to protect: CCA works to keep nearby points close in the reduced output you actually see, while Sammon's works to keep nearby points close in the original data space. This matters because choosing one over the other changes whether you trust the output to reflect true local structure in the result or to faithfully mirror the input's local geometry.