Closed exalate-issue-sync[bot] closed 1 year ago
Former user commented: The R function is now available as h2o.proj_archetypes(myGLRMmodel, myGLRMtrain) with an optional reverse_transform = TRUE/FALSE parameter, indicating whether the final projection should be re-scaled (assuming we transformed the training data with say "DESCALE" when building the GLRM model). The Python function is basically identical, myGLRMmodel.proj_archetypes(myGLRMtrain).
JIRA Issue Migration Info
Jira Issue: PUBDEV-2368 Assignee: Former user Reporter: Former user State: Resolved Fix Version: N/A Attachments: N/A Development PRs: N/A
For non-numeric training features, an additional minimization step must be done to project the archetypes back into the "closest" feature. This is equivalent to doing the reconstruction (prediction) with X consisting of indicator columns, i.e. [[1,0,0,...,0], [0,1,0,...,0], ..., [0,0,0,...,1]]