The use of abstraction has been largely informal. As a consequence, it has often been difficult to see how or why a particular abstraction works. This paper attempts to help correct this trend by presenting a formal theory of abstraction. We use this theory to characterise the different types of abstraction that can be built; the different classes of abstractions we identify capture the majority of abstractions of which we are aware. We end by proposing a method for automatically building one very common type of abstraction, that used in Abstrips; our proposal is motivated by consideration of the various formal properties that such a method should possess.
|Number of pages||10|
|Journal||Automatic Generation of Approximations and Abstactions|
|Publication status||Published - 1990|