The Sammon Mapping Module

This module implements the Sammon Mapping algorithm for non-linear dimensionality reduction. The implementation used is based off a Matlab version given by Gavin C. Cawley, the source of which can be found here.

Essentially, the algorithm works by defining a metric on high-dimensional source data and then seeking to reconstruct a set of low-dimensional projected data such that the pairwise distances given by this metric are well-preserved by the projection metric.

---

sammon(data, target_dim=2, max_iterations=250, max_halves=10)

• data: The source data array in which rows are patters and columns are features.
• target_dim: The dimension of the target projection space.
• max_iterations: The maximum number of iterations of the algorithm to run.
• max_halves: The maximum number of step halvings.