ISOMAP is one of the approaches for manifold learning, which generalizes linear framework to nonlinear data structures. Similar to MDS, ISOMAP creates a visual presentation of similarities or dissimilarities (distance) of a number of objects. However, as the data is structured in a nonlinear format, the Euclidian distance measure of MDS is replaced by the geodesic distance of a data manifold in ISOMAP. In this recipe, we will illustrate how to perform a nonlinear dimension reduction with ISOMAP.
In this recipe, we will use the
digits data from
RnavGraphImageData as our input source.
Perform the following steps to perform nonlinear dimension reduction with ISOMAP: