Encoding Class Repair and Prediction

Class Tree
Description This class contains encodings described in the following KR'10 paper:

Martin Gebser, Carito Guziolowski, Mihail Ivanchev, Torsten Schaub, Anne Siegel, Sven Thiele and Philippe Veber. Repair and Prediction (under Inconsistency) in Large Biological Networks with Answer Set Programming. In Proceedings of the Twelfth International Conference on Principles of Knowledge Representation and Reasoning (KR'10), AAAI Press, 2010.



ENCODINGS
=========

- input.gringo : automatic input detection (basic program)
- repair_config.gringo : repair type configuration (basic program)
- repair_core.gringo : candidate repair generation (extended program)
- repair_cardinality.gringo : cardinality minimization (optimize program)
- repair_subset.gringo : subset minimization (disjunctive program)


REQUIRED/RECOMMENDED PROGRAMS
=============================

- gringo : version 2.0.2 (or later)
- clasp : version 1.2.0 (or later)
- claspD : version 1.1.0 (or later)

Sources (and binaries) are available at: http://potassco.sourceforge.net/.


USAGE
=====

- Configuration of repair types

Admissible repair operations are specified by providing a value for the
constant "repair". Thereby, the following letters stand for repair types:

* 'a' : add edges (model repair)
* 'e' : flip signs of edges (model repair)
* 'g' : globally make vertices input (model repair)
* 'i' : make vertices input for an experiment (data repair)
* 'v' : flip signs of vertices (data repair)

Any string composed of these letters in alphabetic order, i.e., a substring
of 'aegiv', determines a collection of admissible repair operations. In the
example calls below, we denote any such configuration string by REPAIR.

For large influence graphs, including 'a' as repair type leads to a
(possibly too) huge ground instantiation.

- Cardinality-minimal repair

Let GRAPH denote a file containing the description of an influence graph (in
terms of facts), and likewise let OBS_1,...,OBS_n contain experimental data.
Then, a cardinality-minimal repair along with the minimum number CARD_MIN of
repair operations to apply are computed using a call of the following form:

gringo -c repair=REPAIR input.gringo repair_config.gringo repair_core.gringo \
repair_cardinality.gringo GRAPH OBS_1 ... OBS_n | clasp 0

Note that further options, e.g., "--opt-heu", "--restart-on-model", etc.,
can be set for clasp.

- Prediction

Prediction w.r.t. cardinality-minimal repairs is accomplished via clasp's
cautious reasoning mode. A respective call is of the following form:

gringo -c repair=REPAIR input.gringo repair_config.gringo repair_core.gringo \
repair_cardinality.gringo GRAPH OBS_1 ... OBS_n | clasp 0 --opt-all \
--opt-value=CARD_MIN --cautious

- Subset-minimal repair

Subset-minimal repairs can be computed using claspD instead of clasp. A
respective call, where NUM stands for the desired number of repairs and
witnessing vertex/edge labelings, is of the following form:

gringo -c repair=REPAIR input.gringo repair_config.gringo repair_core.gringo \
repair_subset.gringo GRAPH OBS_1 ... OBS_n | claspD NUM


FURTHER INFORMATION
===================

For further details, especially on the logical background of repair and
on encodings, see the aforementioned KR'10 paper or consult:
http://www.cs.uni-potsdam.de/bioasp/.
Requests by e-mail can be sent to authors of the aforementioned KR'10 paper.
Encodings
1 - 5 of 5