Label Prediction

  • Implements a family of methods for predicting labels for an input. In general, we assume inputs in Rd , though class X is maintained abstract.
  • L is the class of labels. We assume that a label does not have internal structure. We abuse notation and denote as L the number of labels.
  • Y = {0,1}L
    • Binary classification, L=1
    • Multiclass classification, constrained to y in Y such that sum { l } y_l =1
    • Multilabel classification, no constraints
  • We may consider a ranking setting, Y = RL