boosting

Terms from Artificial Intelligence: humans at the heart of algorithms

Page numbers are for draft copy at present; they will be replaced with correct numbers when final book is formatted. Chapter numbers are correct and will not change now.

Boosting is the general term for improving an initially poor or weak solution and maijg it better. For example, rone can take a number of different classifiers, each of which is not very accurate on theor own, and combaine them by giving each a "vote" in the final outcome, whuch may be far more accurate than any single classifier on its own. A different form of boosting can be to identify training itms that are consistenly classified poorly and addng weight to them during successive trainining rounds.

Used on Chap. 8: page 161; Chap. 16: page 387; Chap. 22: pages 546, 547; page xxvii