Transductive Conformal Predictor

In basic conformal prediction the goal is to predict the label of a test object {⚠ $x_{n+1}$} given a training set {⚠ $z_1,\ldots,z_n$}. In transductive conformal prediction, we are given a set {⚠ $x_{n+1},\ldots,x_{n+k}$} of test objects and the goal is to predict their labels without getting any intermediate feedback. This is motivated by the general problem of transduction in machine learning.

References

  • Vladimir Vapnik (2000). The Nature of Statistical Learning Theory, 2nd edition. Springer, New York. Transductive inference is discussed in Chapter 9.
  • Vladimir Vovk (2013). Transductive conformal predictors, On-line Compression Modelling Project (New Series), Working Paper 8.