eduzhai > Applied Sciences > Computer Science >

Focused Bayesian Prediction

  • KanKan
  • (0) Download
  • 20210425
  • Save

... pages left unread,continue reading

Document pages: 54 pages

Abstract: We propose a new method for conducting Bayesian prediction that deliversaccurate predictions without correctly specifying the unknown true datagenerating process. A prior is defined over a class of plausible predictivemodels. After observing data, we update the prior to a posterior over thesemodels, via a criterion that captures a user-specified measure of predictiveaccuracy. Under regularity, this update yields posterior concentration onto theelement of the predictive class that maximizes the expectation of the accuracymeasure. In a series of simulation experiments and empirical examples we findnotable gains in predictive accuracy relative to conventional likelihood-basedprediction.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...