Inductive programming: Difference between revisions

Revision as of 16:47, 27 August 2015

Revision as of 20:12, 29 August 2015

Line 20:

These approaches were split into two phases: first, input-output examples are transformed into non-recursive programs (traces) using a small set of basic operators; second, regularities in the traces are searched for and used to fold them into a recursive program. The main results until the mid 1980s are surveyed by Smith.<ref>{{cite journal|first1=D.R.|last1=Smith|title=The synthesis of LISP programs from examples: a survey|editor1-first=A.W.|editor1-last=Biermann|editor2-first=G.|editor2-last=Guiho|publisher=Macmillan|journal=Automatic program construction techniques|pages=307–324|year=1984}}</ref> Due to limited progress with respect to the range of programs that could be synthesized, research activities decreased significantly in the next decade.

The advent of logic programming brought a new elan but also a new direction in the early 1980s, especially due to the MIS system of Shapiro<ref>{{cite book|first1=E.Y.|last1=Shapiro|title=Algorithmic program debugging|publisher=The MIT Press|year=1983}}</ref> eventually spawning the new field of inductive logic programming (ILP).<ref>{{citeCite journal doi| last1 = Muggleton | first1 = S. | title = Inductive logic programming | doi = 10.1007/BF03037089 | journal = New Generation Computing | volume = 8 | issue = 4 | pages = 295–318 | year = 1991 | pmid = | pmc = }}</ref> The early works of Plotkin,<ref>{{cite journal|first1=Gordon D.|last1=Plotkin|title=A Note on Inductive Generalization|editor1-first=B.|editor1-last=Meltzer|editor2-first=D.|editor2-last=Michie|publisher=Edinburgh University Press|journal=Machine Intelligence|volume=5|pages=153–163|year=1970}}</ref><ref>{{cite journal|first1=Gordon D.|last1=Plotkin|title=A Further Note on Inductive Generalization|editor1-first=B.|editor1-last=Meltzer|editor2-first=D.|editor2-last=Michie|publisher=Edinburgh University Press|journal=Machine Intelligence|volume=6|pages=101–124|year=1971}}</ref> and his "''relative least general generalization (rlgg)''", had an enormous impact in inductive logic programming. Most of ILP work addresses a wider class of problems, as the focus is not only on recursive logic programs but on machine learning of symbolic hypotheses from logical representations. However, there were some encouraging results on learning recursive Prolog programs such as quicksort from examples together with suitable background knowledge, for example with GOLEM.<ref>{{cite journal|first1=S.H.|last1=Muggleton|first2=C.|last2=Feng|title=Efficient induction of logic programs|publisher=the Japanese Society for AI|journal=Proceedings of the Workshop on Algorithmic Learning Theory|volume=6|pages=368–381|year=1990}}</ref> But again, after initial success, the community got disappointed by limited progress about the induction of recursive programs<ref>{{cite journal|

first1=J.R.|last1=Quinlan|first2=R.M.|last2=Cameron-Jones|

title=Avoiding Pitfalls When Learning Recursive Theories|

Line 276:

* {{cite journal|first1=M.|last1=Hofmann|first2=E.|last2=Kitzelmann|title=A unifying framework for analysis and evaluation of inductive programming systems|journal=Proceedings of the Second Conference on Artificial General Intelligence|pages=55–60|year=2009}}

* {{Cite journal | last1 = Muggleton | first1 = S. | last2 = De Raedt | doi = 10.1016/0743-1066(94)90035-3 | first2 = L. | title = Inductive Logic Programming: Theory and methods | journal = The Journal of Logic Programming | volume = 19-20 | pages = 629–679 | year = 1994 | pmid = | pmc = }}

* {{cite doi|10.1016/0743-1066(94)90035-3}}

* {{cite book | first1 = N. | last1 = Lavrac | first2 = S. | last2 = Dzeroski | title = Inductive Logic Programming: Techniques and Applications | publisher = Ellis Horwood | ___location = New York | year = 1994 | isbn = 0-13-457870-8 | url = http://www-ai.ijs.si/SasoDzeroski/ILPBook/ }}