Skip to content

Latest commit

 

History

History
26 lines (21 loc) · 654 Bytes

todo.md

File metadata and controls

26 lines (21 loc) · 654 Bytes

To Do

  • Make multidimensional categorical hinge loss

  • Set up a single command to: a) pick m random subsets of data, b) choose model of rank k, regularization constant α for each subset, c) impute missing data from each of the m selected models.

    • (nandana will do this)
  • Call julia from R

    • i write a wrapper
    • nandana writes the UX
  • Documentation!

    • how to think about mpca
    • imputation
    • error metrics for cross validation
    • new syntax for fitting data frame
      • how to specify loss function(s)
    • parallel fitting
    • full rank model / prisma
  • Poisson loss

    • scaling?
    • to log or not to log? that is the interpretative issue