A target function, in machine learning, is a method for solving a problem that an AI algorithm parses its training data to find. Once an algorithm finds its target function, that function can be used to predict results (predictive analysis). The function can then be used to find output data related to inputs for real problems where, unlike training sets, outputs are not included.
The target function is essentially the formula that an algorithm feeds data to in order to calculate predictions. As in algebra, it is common when training AI to find the variable from the solution, working in reverse. The function as defined by f is applied to the input (I) to produce the output (I), Therefore O= f(I).
Analyzing the massive amounts of data related to its given problem, an AI derives understanding of previously unspecified rules by detecting consistencies in the data. The observations of inherent rules about how the studied subject operates inform the AI on how to process future data that does not include an output by applying this previously unknown function.