Searching for a record in a data set is a commonplace operation in data processing and analysis. When the match to the target is exact, the operation is straightforward, but many searches must be inexact, for example, searching for similar faces, or searching for similar crimes. We call this kind of search fuzzy, not in the mathematical sense as it is used in fuzzy logic, but in the everyday sense of inexact. When this kind of fuzzy searching is performed using a neural network, we call it neuro-fuzzy searching.
Neuro-fuzzy searching is accomplished by training a neural network model to recognize the target, the object of the search, and produce a score that rates the similarity of an example to the target. This model is then used to score the database to be searched, and we can then select the example or examples that are most similar to the target.