An objective function is a measure of how similar a prediction of a value and the actual value are. Usually, we are looking to find the set of parameters that lead to the smallest possible cost which would imply that your algorithm will perform well.
F. Huszár, and D. Duvenaud. (2012)cite arxiv:1204.1664Comment: Accepted as an oral presentation at Uncertainty in Artificial Intelligence 2012. Updated to fix several typos.
C. Elliott. (2018)cite arxiv:1804.00746Comment: 37 pages with proof appendices and 15 figures. Extended version of a paper appearing at ICFP 2018. More info at http://conal.net/papers/essence-of-ad/.