eduzhai > Applied Sciences > Computer Science >

A Hierarchy of Limitations in Machine Learning

  • KanKan
  • (0) Download
  • 20210424
  • Save

... pages left unread,continue reading

Document pages: 68 pages

Abstract: "All models are wrong, but some are useful ", wrote George E. P. Box (1979).Machine learning has focused on the usefulness of probability models forprediction in social systems, but is only now coming to grips with the ways inwhich these models are wrong---and the consequences of those shortcomings. Thispaper attempts a comprehensive, structured overview of the specific conceptual,procedural, and statistical limitations of models in machine learning whenapplied to society. Machine learning modelers themselves can use the describedhierarchy to identify possible failure points and think through how to addressthem, and consumers of machine learning models can know what to question whenconfronted with the decision about if, where, and how to apply machinelearning. The limitations go from commitments inherent in quantificationitself, through to showing how unmodeled dependencies can lead tocross-validation being overly optimistic as a way of assessing modelperformance.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×