SlideShare a Scribd company logo

UnBBayes Overview

•
Rommel Carvalho, profile picture
Rommel Carvalho

The document provides an overview of Unbbayes, a system developed by Rommel Novaes Carvalho and his team for working with Bayesian networks and probabilistic ontologies. It highlights the evolution of Unbbayes, its architecture, methodologies, and applications, particularly focusing on the integration of semantic web technologies. The presentation also discusses the motivation behind its development, including the need to exploit data from disparate sources and manage uncertainty in semantic contexts.

Education
1 of 116
Downloaded 139 times
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes Overview Slides Potpourri Rommel Novaes Carvalho and GIA (Artificial Intelligence Group) from UnB GMU - September 19th 2008
©2008 Rommel Novaes Carvalho – University of Brasília Team & Background Advisor: Dr. Marcelo Ladeira - UnB Co-advisor: Dr. Paulo C. G. Costa - GMU Bachelor Degree in CS: Laecio L. Santos - UnB Shou Matsumoto - UnB Consultant: Dr. Kathryn B. Laskey - GMU Papers IADIS FLAIRS ISDA - IEEE Selected to extend it as a book chapter 2
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor XMLBIF 0.4 (never published, though) 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor XMLBIF 0.4 (never published, though) UnBBayes Server (J2EE) 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor XMLBIF 0.4 (never published, though) UnBBayes Server (J2EE) MEBN/PR-OWL 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor XMLBIF 0.4 (never published, though) UnBBayes Server (J2EE) MEBN/PR-OWL Other things Gibbs for missing values Monte Carlo Etc (not even I know it all!) UnBMiner (maybe some other day...) People involved > 15 3
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes before MEBN BN, ID, MSBN, and UnBBayes Server
©2008 Rommel Novaes Carvalho – University of Brasília Agenda I BN ID MSBN UnBBayes Server 5
©2008 Rommel Novaes Carvalho – University of Brasília BN Asia BN - ID - MSBN - UnBBayes Server 6
©2008 Rommel Novaes Carvalho – University of Brasília BN - Compile Asia BN - ID - MSBN - UnBBayes Server 7
©2008 Rommel Novaes Carvalho – University of Brasília BN - Update Beliefs Asia BN - ID - MSBN - UnBBayes Server 8
©2008 Rommel Novaes Carvalho – University of Brasília ID Car Buyer BN - ID - MSBN - UnBBayes Server 9
©2008 Rommel Novaes Carvalho – University of Brasília ID - Compile Car Buyer BN - ID - MSBN - UnBBayes Server 10
©2008 Rommel Novaes Carvalho – University of Brasília ID - Update Beliefs Car Buyer BN - ID - MSBN - UnBBayes Server 11
©2008 Rommel Novaes Carvalho – University of Brasília MSBN Extended Asia BN - ID - MSBN - UnBBayes Server 12
©2008 Rommel Novaes Carvalho – University of Brasília MSBN - Compile Extended Asia BN - ID - MSBN - UnBBayes Server 13
©2008 Rommel Novaes Carvalho – University of Brasília MSBN - Update Beliefs Extended Asia BN - ID - MSBN - UnBBayes Server 14
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes Server CRUD for BN models and evidence history + online reasoner BN - ID - MSBN - UnBBayes Server 15
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes after MEBN
©2008 Rommel Novaes Carvalho – University of Brasília Agenda II Motivation Methodology Understanding MEBN Building a MTheory - Star Trek SSBN in UnBBayes Conclusions 17
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Semantic Web - SW emerges as a motivation for defining: common formats to integrate and combine data formalisms for recording how data relates to real world objects Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Semantic Web - SW emerges as a motivation for defining: common formats to integrate and combine data formalisms for recording how data relates to real world objects Ontologies have precisely defined concepts to represent a certain domain Key for semantic interoperability [Costa et al, 2006] Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Semantic Web - SW emerges as a motivation for defining: common formats to integrate and combine data formalisms for recording how data relates to real world objects Ontologies have precisely defined concepts to represent a certain domain Key for semantic interoperability [Costa et al, 2006] “Washington” - syntatic <> SW <> uncertainty in SW Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Semantic Web - SW emerges as a motivation for defining: common formats to integrate and combine data formalisms for recording how data relates to real world objects Ontologies have precisely defined concepts to represent a certain domain Key for semantic interoperability [Costa et al, 2006] “Washington” - syntatic <> SW <> uncertainty in SW URW3-XG WWW uncertainty reasoning use cases and standard approach Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Uncertainty in the SW Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 19
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Uncertainty in the SW BN [Pearl, 1988] is one of the most promising approaches for dealing with uncertainty in the SW Probability: well fundamented principles and known semantics Limitations in the SW The number of variables has to be known in advance Lack of support to recursive definition Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 19
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Uncertainty in the SW BN [Pearl, 1988] is one of the most promising approaches for dealing with uncertainty in the SW Probability: well fundamented principles and known semantics Limitations in the SW The number of variables has to be known in advance Lack of support to recursive definition Use of the FOL expressiveness to overcome these limitations Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 19
©2008 Rommel Novaes Carvalho – University of Brasília Motivation Uncertainty in the SW BN [Pearl, 1988] is one of the most promising approaches for dealing with uncertainty in the SW Probability: well fundamented principles and known semantics Limitations in the SW The number of variables has to be known in advance Lack of support to recursive definition Use of the FOL expressiveness to overcome these limitations Costa (2005) proposed a First-Order Bayesian framework to probabilistic ontologies based in PR-OWL and MEBN There was no implementation of PR-OWL and MEBN Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 19
©2008 Rommel Novaes Carvalho – University of Brasília Methodology Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Extension for UnBBayes Development in Java with GNU GPL v3 and internationalization Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Extension for UnBBayes Development in Java with GNU GPL v3 and internationalization Selection of free tools FOL: PowerLoom OWL: Protégé Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Extension for UnBBayes Development in Java with GNU GPL v3 and internationalization Selection of free tools FOL: PowerLoom OWL: Protégé Friendly GUI Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Extension for UnBBayes Development in Java with GNU GPL v3 and internationalization Selection of free tools FOL: PowerLoom OWL: Protégé Friendly GUI Experimental evaluation through a toy use case Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
©2008 Rommel Novaes Carvalho – University of Brasília Understanding MEBN BN + FOL = MEBN The knowledge is represented as a set of MEBN fragments (MFrags, for short) organized as a MEBN theory (MTheories, for short) Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 21
©2008 Rommel Novaes Carvalho – University of Brasília Understanding MEBN MTheory and MFrag MEBN theories extend ordinary Bayesian networks to provide an inner structure for RVs that take arguments that refer to entities in the domain of application. A MEBN theory implicitly expresses a JPD over truth-values of sets of FOL sentences. Context Nodes Input Nodes Resident Nodes Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 22
©2008 Rommel Novaes Carvalho – University of Brasília Understanding MEBN MEBN + OWL = PR-OWL PR-OWL was proposed by Costa (2005) as an extension to the OWL language, based in MEBN, that enables a probabilistic distribution over over models of any FOL axiomatized theory. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 23
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes Architecture Version Layers Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 24
©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes Architecture Extension Points Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 25
©2008 Rommel Novaes Carvalho – University of Brasília MEBN before UnBBayes Protégé GUI Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 26
©2008 Rommel Novaes Carvalho – University of Brasília MEBN after UnBBayes UnBBayes GUI Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 27
©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Star Trek Model Starship Zone TimeStep [ord] SensorReport Sensor CloakMode Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 28
©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Resident node L OW P R- L OW PR- Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 29
©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Input node E BN M L OW BN P R- M E E BN M Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 30
©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Context node Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 31
©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Proposed grammar for dynamic CPT E BN M ithf w i w ted noes n Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 32
©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Saving ubf and pr-owl files UnBBayes Save UBF PR-OWL Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 33
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Inserting entities in the KB Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 34
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Inserting evidences in the KB Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 35
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- LIST), search for evidence in the KB. If found it, return it. If not, continue. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- LIST), search for evidence in the KB. If found it, return it. If not, continue. ii. Search for the resident node that has the name NODE and get its MFrag. Once NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- LIST), search for evidence in the KB. If found it, return it. If not, continue. ii. Search for the resident node that has the name NODE and get its MFrag. Once NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. iii. Verify in the KB which context node refers to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. ii. Search for the resident node that has the name NODE and get its MFrag. Once NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. iii. Verify in the KB which context node refers to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. v. For each father of NODE, go to step (i), ii. Search for the resident node that has the replacing the OVs by the known entities name NODE and get its MFrag. Once (contained in the query or KB). NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. iii. Verify in the KB which context node refers to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. v. For each father of NODE, go to step (i), ii. Search for the resident node that has the replacing the OVs by the known entities name NODE and get its MFrag. Once (contained in the query or KB). NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. vi. Create the NODE’s CPT. iii. Verify in the KB which context node refers to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. v. For each father of NODE, go to step (i), ii. Search for the resident node that has the replacing the OVs by the known entities name NODE and get its MFrag. Once (contained in the query or KB). NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. vi. Create the NODE’s CPT. iii. Verify in the KB which context node refers vii. Finish. to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. v. For each father of NODE, go to step (i), ii. Search for the resident node that has the replacing the OVs by the known entities name NODE and get its MFrag. Once (contained in the query or KB). NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. vi. Create the NODE’s CPT. iii. Verify in the KB which context node refers vii. Finish. to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, BN mark the MFrag to use the default ME distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Query HarmPotential(!ST4, !T3) = ? Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 37
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK OK OK OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 38
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 39
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 40
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K GE OK O AN 2R OK S ER P HA ] )= [F ,! T0 T4 PT (!S a teC FO er D g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 41
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK C PT ate er g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 42
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK C PT ate er g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 43
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 44
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 45
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O ] [F Z2 z =! Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 46
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 47
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK C PT r ate e ne g Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 48
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK C PT r ate e ne g Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 49
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O ] [F Z2 z =! Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 50
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 51
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK C PT ate er g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 52
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O C PT r ate e ne g Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 53
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O C PT ate er g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 54
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O PT OK a teC n er ge Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 55
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 56
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes BN inference Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 57
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes BN inference Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 58
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes BN inference Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 59
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Explosive* (goal driven with evidence below)‫‏‬ SSBN topology 1 3 Does NODE has children RNs in the same MFrag? If 1 Evidence for yes, call 1 for each NODE in KB? If CHILD. yes, finish. 4 Get related INs to 2 Get related NODE. If INs have context nodes children, call 1 for and evaluate each CHILD. them. OK OK OK OK 5 If NODE is permanent, evaluate NODE’s parents. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 60
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Explosive* (goal driven with evidence below)‫‏‬ Permanent Nodes 1 1 The query node is 1 4 always permanent. 2 Parents from a 2 3 5 permanent node are also permanent (except for evidence nodes). 3 All findings are permanent nodes. 1 6 4 Parents are only evaluated if node is permanent. 2 3 5 Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 61
©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Explosive* (goal driven with evidence below)‫‏‬ General algorithm 1 Finding No parent 1 Generate SSBN topology. 2 Create CPT for permanent nodes. 3 Remove temporary nodes. 4 Compile the network. 5 Set evidences and update beliefs. Finding below Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 62
©2008 Rommel Novaes Carvalho – University of Brasília General Metaphor Simple use of BN for end users Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 63
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; 2. Identify the entities present in the domain. If you are having trouble trying to come up with different entities, you probably do not need to use MEBN. Try using just BN, instead; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; 2. Identify the entities present in the domain. If you are having trouble trying to come up with different entities, you probably do not need to use MEBN. Try using just BN, instead; 3. Identify logical groups (there are groups of information that are or could be logically put together) for identifying sets of entities; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; 2. Identify the entities present in the domain. If you are having trouble trying to come up with different entities, you probably do not need to use MEBN. Try using just BN, instead; 3. Identify logical groups (there are groups of information that are or could be logically put together) for identifying sets of entities; 4. Identify criterias that can classify in some way the identified entities. This will help you choose which entities are relevant and which are not to solve the problem (discard them from your model). This step can also help to detect uncertain about the existence of information and to identify contexts where certain informations can be considered valid; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; 2. Identify the entities present in the domain. If you are having trouble trying to come up with different entities, you probably do not need to use MEBN. Try using just BN, instead; 3. Identify logical groups (there are groups of information that are or could be logically put together) for identifying sets of entities; 4. Identify criterias that can classify in some way the identified entities. This will help you choose which entities are relevant and which are not to solve the problem (discard them from your model). This step can also help to detect uncertain about the existence of information and to identify contexts where certain informations can be considered valid; 5. Identify the attributes the entities can have; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; 8. Evaluate if MEBN is necessary and/or sufficient for modeling the problem; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; 8. Evaluate if MEBN is necessary and/or sufficient for modeling the problem; 9. Evaluate if the identified entities are enough for your MEBN model; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; 8. Evaluate if MEBN is necessary and/or sufficient for modeling the problem; 9. Evaluate if the identified entities are enough for your MEBN model; 10. Map the entities, groups, rules, and relations identified to their respective MEBN element (MFrag, Node, etc); Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; 8. Evaluate if MEBN is necessary and/or sufficient for modeling the problem; 9. Evaluate if the identified entities are enough for your MEBN model; 10. Map the entities, groups, rules, and relations identified to their respective MEBN element (MFrag, Node, etc); 11. Design the model in UnBBayes. Here, you might need to change your model a little bit, because UnBBayes has some singularities due to implementation (the way it implements recursion, possible states for a resident node, etc). Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 66
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Results Contributions to PR-OWL/MEBN Tool for representing and reasoning in probabilistic ontologies Complete use case example Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 66
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Results Contributions to PR-OWL/MEBN Tool for representing and reasoning in probabilistic ontologies Complete use case example Scientific contributions PR-OWL Entity as a possible state for a node Global exclusivity Built-in recursion MEBN New algorithm for creating SSBN Grammar for creating dynamic CPT Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 66
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 67
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Technological contributions New algorithm for creating SSBN implemented First implementation of PR-OWL/MEBN in the world GNU GPL v3, open source and free www.sourceforge.net/projects/unbbayes Platform independent – Java Internationalization OWL compatible Friendly GUI Compiler for dynamic CPT Format .ubf Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 67
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 68
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Limitations Query with only one node No recursion explicit stop condition GUI Node’s size is not proportional to its label Edges do not get to the node’s boundary No overall MTheory visualization No MFrag reuse Some elements present in Laskey 2007 were not implemented Save finding in PR-OWL Built-in recursion in PR-OWL “Likely” and “FOL” findings Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 68
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 69
©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Future work Real use case Transparency and corruption prevention CGU - Brazilian General Comptroller Office ONR? Implement limitations Integrate OWL reasoner with MEBN reasoner => real PR-OWL reasoner MCMC (Gibbs) for approximate inference OOBN ...? Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 69
©2008 Rommel Novaes Carvalho – University of Brasília Obrigado! Slides Potpourri Rommel Novaes Carvalho and GIA (Artificial Intelligence Group) from UnB GMU - September 19th 2008

More Related Content

PDF
UnBBayes Plugin Framework
Rommel Carvalho
 
PDF
Web ARChive (WARC) File Format
Sawood Alam
 
PPSX
Modelo OSI
Ricardo Batista
 
PDF
Insights on the configuration and performances of SOME/IP Service Discovery
Nicolas Navet
 
PPTX
Personal kompüterin baza apparat hissəsi. Çipset və ana plataları..pptx
JafarAgabalayev1
 
PPTX
What is rtmp
StreamHash
 
DOCX
Distributed system unit II according to syllabus of RGPV, Bhopal
NANDINI SHARMA
 
PPTX
Seguranca de rede
Universidade Federal do Pampa
 
UnBBayes Plugin Framework
Rommel Carvalho
 
Web ARChive (WARC) File Format
Sawood Alam
 
Modelo OSI
Ricardo Batista
 
Insights on the configuration and performances of SOME/IP Service Discovery
Nicolas Navet
 
Personal kompüterin baza apparat hissəsi. Çipset və ana plataları..pptx
JafarAgabalayev1
 
What is rtmp
StreamHash
 
Distributed system unit II according to syllabus of RGPV, Bhopal
NANDINI SHARMA
 
Seguranca de rede
Universidade Federal do Pampa
 

Viewers also liked (7)

PDF
Probabilistic Ontology: Representation and Modeling Methodology
Rommel Carvalho
 
PDF
UnBBayes-PRM - On Implementing Probabilistic Relational Models
Rommel Carvalho
 
PDF
Brasilia
Lucas Rizoli
 
PPTX
1.2. victor andrade
Plataforma Tecnológica da Bicicleta e Mobilidade Suave
 
PPTX
Planning in Brasilia
Gayathri Kumari
 
PDF
Presentation Brasilia
ivanleung
 
PPTX
Brasilia town planning
Shubhangi Bhatnagar
 
Probabilistic Ontology: Representation and Modeling Methodology
Rommel Carvalho
 
UnBBayes-PRM - On Implementing Probabilistic Relational Models
Rommel Carvalho
 
Brasilia
Lucas Rizoli
 
1.2. victor andrade
Plataforma Tecnológica da Bicicleta e Mobilidade Suave
 
Planning in Brasilia
Gayathri Kumari
 
Presentation Brasilia
ivanleung
 
Brasilia town planning
Shubhangi Bhatnagar
 
Ad

Similar to UnBBayes Overview (20)

DOCX
Bioinformatics Final Report
Shruthi Choudary
 
PDF
Intelligible Machine Learning with Malibu for bioinformatics ...
butest
 
PPTX
Lecture_1_Introduction_Bioinformatics.pptx
90loiq2y9
 
PDF
Bioinformatics Meets Information Retrieval: State of the Art and a Case Study
Eloisa Vargiu
 
PDF
Producing, Publishing and Consuming Linked Data Three lessons from the Bio2RD...
François Belleau
 
PPTX
Cloud bioinformatics 2
ARPUTHA SELVARAJ A
 
PDF
Introduction - Lecture 1 - Seminar Web Information Systems Technology (WE-DIN...
Beat Signer
 
PDF
Javatech An Introduction To Scientific And Technical Computing With Java Clar...
papaivierkiy
 
PDF
Initial steps towards a production platform for DNA sequence analysis on the ...
Barbera van Schaik
 
PDF
Datos enlazados BNE and MARiMbA
Daniel Vila Suero
 
PPTX
SMWCon 2012 Linked Data Visualizations
William Smith
 
PDF
A Web Services Based Framework For Uniform Integration Of Command-Line Bioinf...
Kim Daniels
 
PPT
A graduate student's experience in bioinformatics
Morgan Langille
 
PPT
Interactive Datamining of Large-Scale Screening Datasets
Frank Oellien
 
PDF
LECTURE NOTES ON BIOINFORMATICS
MSCW Mysore
 
PPTX
BioInformatics Software
university of education,Lahore
 
ODP
Simplify the complexity of your business processes
Kris Verlaenen
 
PDF
Bioinformatics مي.pdf
nedalalazzwy
 
PDF
Publishing linked data from relational databases
Iván Ruiz-Rube
 
PDF
Processing introduction
Sam Agten
 
Bioinformatics Final Report
Shruthi Choudary
 
Intelligible Machine Learning with Malibu for bioinformatics ...
butest
 
Lecture_1_Introduction_Bioinformatics.pptx
90loiq2y9
 
Bioinformatics Meets Information Retrieval: State of the Art and a Case Study
Eloisa Vargiu
 
Producing, Publishing and Consuming Linked Data Three lessons from the Bio2RD...
François Belleau
 
Cloud bioinformatics 2
ARPUTHA SELVARAJ A
 
Introduction - Lecture 1 - Seminar Web Information Systems Technology (WE-DIN...
Beat Signer
 
Javatech An Introduction To Scientific And Technical Computing With Java Clar...
papaivierkiy
 
Initial steps towards a production platform for DNA sequence analysis on the ...
Barbera van Schaik
 
Datos enlazados BNE and MARiMbA
Daniel Vila Suero
 
SMWCon 2012 Linked Data Visualizations
William Smith
 
A Web Services Based Framework For Uniform Integration Of Command-Line Bioinf...
Kim Daniels
 
A graduate student's experience in bioinformatics
Morgan Langille
 
Interactive Datamining of Large-Scale Screening Datasets
Frank Oellien
 
LECTURE NOTES ON BIOINFORMATICS
MSCW Mysore
 
BioInformatics Software
university of education,Lahore
 
Simplify the complexity of your business processes
Kris Verlaenen
 
Bioinformatics مي.pdf
nedalalazzwy
 
Publishing linked data from relational databases
Iván Ruiz-Rube
 
Processing introduction
Sam Agten
 
Ad

More from Rommel Carvalho (20)

PPTX
Ouvidoria de Balcão vs Ouvidoria Digital: Desafios na Era Big Data
Rommel Carvalho
 
PDF
Como transformar servidores em cientistas de dados e diminuir a distância ent...
Rommel Carvalho
 
PPTX
Proposta de Modelo de Classificação de Riscos de Contratos Públicos
Rommel Carvalho
 
PPTX
Categorização de achados em auditorias de TI com modelos supervisionados e nã...
Rommel Carvalho
 
PPTX
Mapeamento de risco de corrupção na administração pública federal
Rommel Carvalho
 
PDF
Ciência de Dados no Combate à Corrupção
Rommel Carvalho
 
PDF
Aplicação de técnicas de mineração de textos para classificação automática de...
Rommel Carvalho
 
PDF
Filiação partidária e risco de corrupção de servidores públicos federais
Rommel Carvalho
 
PDF
Uso de mineração de dados e textos para cálculo de preços de referência em co...
Rommel Carvalho
 
PDF
Detecção preventiva de fracionamento de compras
Rommel Carvalho
 
PDF
Identificação automática de tipos de pedidos mais frequentes da LAI
Rommel Carvalho
 
PDF
BMAW 2014 - Using Bayesian Networks to Identify and Prevent Split Purchases i...
Rommel Carvalho
 
PDF
A GUI for MLN
Rommel Carvalho
 
PDF
URSW 2013 - UMP-ST plug-in
Rommel Carvalho
 
PDF
Integração do Portal da Copa @ Comissão CMA do Senado Federal
Rommel Carvalho
 
KEY
Dados Abertos Governamentais
Rommel Carvalho
 
KEY
Modeling a Probabilistic Ontology for Maritime Domain Awareness
Rommel Carvalho
 
PDF
SWRL-F - A Fuzzy Logic Extension of the Semantic Web Rule Language
Rommel Carvalho
 
ODP
Default Logics for Plausible Reasoning with Controversial Axioms
Rommel Carvalho
 
PDF
Tractability of the Crisp Representations of Tractable Fuzzy Description Logics
Rommel Carvalho
 
Ouvidoria de Balcão vs Ouvidoria Digital: Desafios na Era Big Data
Rommel Carvalho
 
Como transformar servidores em cientistas de dados e diminuir a distância ent...
Rommel Carvalho
 
Proposta de Modelo de Classificação de Riscos de Contratos Públicos
Rommel Carvalho
 
Categorização de achados em auditorias de TI com modelos supervisionados e nã...
Rommel Carvalho
 
Mapeamento de risco de corrupção na administração pública federal
Rommel Carvalho
 
Ciência de Dados no Combate à Corrupção
Rommel Carvalho
 
Aplicação de técnicas de mineração de textos para classificação automática de...
Rommel Carvalho
 
Filiação partidária e risco de corrupção de servidores públicos federais
Rommel Carvalho
 
Uso de mineração de dados e textos para cálculo de preços de referência em co...
Rommel Carvalho
 
Detecção preventiva de fracionamento de compras
Rommel Carvalho
 
Identificação automática de tipos de pedidos mais frequentes da LAI
Rommel Carvalho
 
BMAW 2014 - Using Bayesian Networks to Identify and Prevent Split Purchases i...
Rommel Carvalho
 
A GUI for MLN
Rommel Carvalho
 
URSW 2013 - UMP-ST plug-in
Rommel Carvalho
 
Integração do Portal da Copa @ Comissão CMA do Senado Federal
Rommel Carvalho
 
Dados Abertos Governamentais
Rommel Carvalho
 
Modeling a Probabilistic Ontology for Maritime Domain Awareness
Rommel Carvalho
 
SWRL-F - A Fuzzy Logic Extension of the Semantic Web Rule Language
Rommel Carvalho
 
Default Logics for Plausible Reasoning with Controversial Axioms
Rommel Carvalho
 
Tractability of the Crisp Representations of Tractable Fuzzy Description Logics
Rommel Carvalho
 

Recently uploaded (20)

PPTX
How to Track Skills & Contracts Using Odoo 18 Employee
Celine George
 
DOCX
Unit 5: Speech-language and swallowing disorders
JELLA VISHNU DURGA PRASAD
 
PDF
Module 2: Public Health History [Tutorial Slides]
JonathanHallett4
 
PPTX
An introduction to Prepositions for beginners.pptx
drsiddhantnagine
 
PDF
Biological Classification Class 11th NCERT CBSE NEET.pdf
NehaRohtagi1
 
PDF
The Minister of Tourism, Culture and Creative Arts, Abla Dzifa Gomashie has e...
nservice241
 
PPTX
CONCEPT OF CHILD CARE. pptx
AneetaSharma15
 
PPTX
INTESTINALPARASITES OR WORM INFESTATIONS.pptx
PRADEEP ABOTHU
 
PPTX
Kanban Cards _ Mass Action in Odoo 18.2 - Odoo Slides
Celine George
 
PPTX
Tips Management in Odoo 18 POS - Odoo Slides
Celine George
 
PDF
Health-The-Ultimate-Treasure (1).pdf/8th class science curiosity /samyans edu...
Sandeep Swamy
 
PPTX
BASICS IN COMPUTER APPLICATIONS - UNIT I
suganthim28
 
PPTX
An introduction to Dialogue writing.pptx
drsiddhantnagine
 
PPTX
HEALTH CARE DELIVERY SYSTEM - UNIT 2 - GNM 3RD YEAR.pptx
Priyanshu Anand
 
PPTX
Care of patients with elImination deviation.pptx
AneetaSharma15
 
PPTX
Cleaning Validation Ppt Pharmaceutical validation
Ms. Ashatai Patil
 
PPTX
How to Close Subscription in Odoo 18 - Odoo Slides
Celine George
 
PDF
Virat Kohli- the Pride of Indian cricket
kushpar147
 
DOCX
Modul Ajar Deep Learning Bahasa Inggris Kelas 11 Terbaru 2025
wahyurestu63
 
PPTX
Artificial Intelligence in Gastroentrology: Advancements and Future Presprec...
AyanHossain
 
How to Track Skills & Contracts Using Odoo 18 Employee
Celine George
 
Unit 5: Speech-language and swallowing disorders
JELLA VISHNU DURGA PRASAD
 
Module 2: Public Health History [Tutorial Slides]
JonathanHallett4
 
An introduction to Prepositions for beginners.pptx
drsiddhantnagine
 
Biological Classification Class 11th NCERT CBSE NEET.pdf
NehaRohtagi1
 
The Minister of Tourism, Culture and Creative Arts, Abla Dzifa Gomashie has e...
nservice241
 
CONCEPT OF CHILD CARE. pptx
AneetaSharma15
 
INTESTINALPARASITES OR WORM INFESTATIONS.pptx
PRADEEP ABOTHU
 
Kanban Cards _ Mass Action in Odoo 18.2 - Odoo Slides
Celine George
 
Tips Management in Odoo 18 POS - Odoo Slides
Celine George
 
Health-The-Ultimate-Treasure (1).pdf/8th class science curiosity /samyans edu...
Sandeep Swamy
 
BASICS IN COMPUTER APPLICATIONS - UNIT I
suganthim28
 
An introduction to Dialogue writing.pptx
drsiddhantnagine
 
HEALTH CARE DELIVERY SYSTEM - UNIT 2 - GNM 3RD YEAR.pptx
Priyanshu Anand
 
Care of patients with elImination deviation.pptx
AneetaSharma15
 
Cleaning Validation Ppt Pharmaceutical validation
Ms. Ashatai Patil
 
How to Close Subscription in Odoo 18 - Odoo Slides
Celine George
 
Virat Kohli- the Pride of Indian cricket
kushpar147
 
Modul Ajar Deep Learning Bahasa Inggris Kelas 11 Terbaru 2025
wahyurestu63
 
Artificial Intelligence in Gastroentrology: Advancements and Future Presprec...
AyanHossain
 

UnBBayes Overview

  • 1. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes Overview Slides Potpourri Rommel Novaes Carvalho and GIA (Artificial Intelligence Group) from UnB GMU - September 19th 2008
  • 2. ©2008 Rommel Novaes Carvalho – University of Brasília Team & Background Advisor: Dr. Marcelo Ladeira - UnB Co-advisor: Dr. Paulo C. G. Costa - GMU Bachelor Degree in CS: Laecio L. Santos - UnB Shou Matsumoto - UnB Consultant: Dr. Kathryn B. Laskey - GMU Papers IADIS FLAIRS ISDA - IEEE Selected to extend it as a book chapter 2
  • 3. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History 3
  • 4. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java 3
  • 5. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN 3
  • 6. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental 3
  • 7. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor 3
  • 8. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor XMLBIF 0.4 (never published, though) 3
  • 9. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor XMLBIF 0.4 (never published, though) UnBBayes Server (J2EE) 3
  • 10. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor XMLBIF 0.4 (never published, though) UnBBayes Server (J2EE) MEBN/PR-OWL 3
  • 11. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes History BN in Delphi ... Migrated to Java BN, ID, and MSBN BN Learning K2, B, CBL-A and CBL-B Incremental Metaphor XMLBIF 0.4 (never published, though) UnBBayes Server (J2EE) MEBN/PR-OWL Other things Gibbs for missing values Monte Carlo Etc (not even I know it all!) UnBMiner (maybe some other day...) People involved > 15 3
  • 12. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes before MEBN BN, ID, MSBN, and UnBBayes Server
  • 13. ©2008 Rommel Novaes Carvalho – University of Brasília Agenda I BN ID MSBN UnBBayes Server 5
  • 14. ©2008 Rommel Novaes Carvalho – University of Brasília BN Asia BN - ID - MSBN - UnBBayes Server 6
  • 15. ©2008 Rommel Novaes Carvalho – University of Brasília BN - Compile Asia BN - ID - MSBN - UnBBayes Server 7
  • 16. ©2008 Rommel Novaes Carvalho – University of Brasília BN - Update Beliefs Asia BN - ID - MSBN - UnBBayes Server 8
  • 17. ©2008 Rommel Novaes Carvalho – University of Brasília ID Car Buyer BN - ID - MSBN - UnBBayes Server 9
  • 18. ©2008 Rommel Novaes Carvalho – University of Brasília ID - Compile Car Buyer BN - ID - MSBN - UnBBayes Server 10
  • 19. ©2008 Rommel Novaes Carvalho – University of Brasília ID - Update Beliefs Car Buyer BN - ID - MSBN - UnBBayes Server 11
  • 20. ©2008 Rommel Novaes Carvalho – University of Brasília MSBN Extended Asia BN - ID - MSBN - UnBBayes Server 12
  • 21. ©2008 Rommel Novaes Carvalho – University of Brasília MSBN - Compile Extended Asia BN - ID - MSBN - UnBBayes Server 13
  • 22. ©2008 Rommel Novaes Carvalho – University of Brasília MSBN - Update Beliefs Extended Asia BN - ID - MSBN - UnBBayes Server 14
  • 23. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes Server CRUD for BN models and evidence history + online reasoner BN - ID - MSBN - UnBBayes Server 15
  • 24. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes after MEBN
  • 25. ©2008 Rommel Novaes Carvalho – University of Brasília Agenda II Motivation Methodology Understanding MEBN Building a MTheory - Star Trek SSBN in UnBBayes Conclusions 17
  • 26. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
  • 27. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
  • 28. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Semantic Web - SW emerges as a motivation for defining: common formats to integrate and combine data formalisms for recording how data relates to real world objects Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
  • 29. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Semantic Web - SW emerges as a motivation for defining: common formats to integrate and combine data formalisms for recording how data relates to real world objects Ontologies have precisely defined concepts to represent a certain domain Key for semantic interoperability [Costa et al, 2006] Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
  • 30. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Semantic Web - SW emerges as a motivation for defining: common formats to integrate and combine data formalisms for recording how data relates to real world objects Ontologies have precisely defined concepts to represent a certain domain Key for semantic interoperability [Costa et al, 2006] “Washington” - syntatic <> SW <> uncertainty in SW Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
  • 31. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Exploitation of data from disparate sources Making semantic information explicit and computationally accessible [Laskey et al, 2007] Semantic Web - SW emerges as a motivation for defining: common formats to integrate and combine data formalisms for recording how data relates to real world objects Ontologies have precisely defined concepts to represent a certain domain Key for semantic interoperability [Costa et al, 2006] “Washington” - syntatic <> SW <> uncertainty in SW URW3-XG WWW uncertainty reasoning use cases and standard approach Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 18
  • 32. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Uncertainty in the SW Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 19
  • 33. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Uncertainty in the SW BN [Pearl, 1988] is one of the most promising approaches for dealing with uncertainty in the SW Probability: well fundamented principles and known semantics Limitations in the SW The number of variables has to be known in advance Lack of support to recursive definition Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 19
  • 34. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Uncertainty in the SW BN [Pearl, 1988] is one of the most promising approaches for dealing with uncertainty in the SW Probability: well fundamented principles and known semantics Limitations in the SW The number of variables has to be known in advance Lack of support to recursive definition Use of the FOL expressiveness to overcome these limitations Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 19
  • 35. ©2008 Rommel Novaes Carvalho – University of Brasília Motivation Uncertainty in the SW BN [Pearl, 1988] is one of the most promising approaches for dealing with uncertainty in the SW Probability: well fundamented principles and known semantics Limitations in the SW The number of variables has to be known in advance Lack of support to recursive definition Use of the FOL expressiveness to overcome these limitations Costa (2005) proposed a First-Order Bayesian framework to probabilistic ontologies based in PR-OWL and MEBN There was no implementation of PR-OWL and MEBN Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 19
  • 36. ©2008 Rommel Novaes Carvalho – University of Brasília Methodology Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
  • 37. ©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
  • 38. ©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
  • 39. ©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Extension for UnBBayes Development in Java with GNU GPL v3 and internationalization Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
  • 40. ©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Extension for UnBBayes Development in Java with GNU GPL v3 and internationalization Selection of free tools FOL: PowerLoom OWL: Protégé Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
  • 41. ©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Extension for UnBBayes Development in Java with GNU GPL v3 and internationalization Selection of free tools FOL: PowerLoom OWL: Protégé Friendly GUI Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
  • 42. ©2008 Rommel Novaes Carvalho – University of Brasília Methodology Discuss formalisms and contributions PR-OWL: Dr. Paulo Cesar G. Costa MEBN: Dr. Kathryn B. Laskey URW3-XG Extension for UnBBayes Development in Java with GNU GPL v3 and internationalization Selection of free tools FOL: PowerLoom OWL: Protégé Friendly GUI Experimental evaluation through a toy use case Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 20
  • 43. ©2008 Rommel Novaes Carvalho – University of Brasília Understanding MEBN BN + FOL = MEBN The knowledge is represented as a set of MEBN fragments (MFrags, for short) organized as a MEBN theory (MTheories, for short) Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 21
  • 44. ©2008 Rommel Novaes Carvalho – University of Brasília Understanding MEBN MTheory and MFrag MEBN theories extend ordinary Bayesian networks to provide an inner structure for RVs that take arguments that refer to entities in the domain of application. A MEBN theory implicitly expresses a JPD over truth-values of sets of FOL sentences. Context Nodes Input Nodes Resident Nodes Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 22
  • 45. ©2008 Rommel Novaes Carvalho – University of Brasília Understanding MEBN MEBN + OWL = PR-OWL PR-OWL was proposed by Costa (2005) as an extension to the OWL language, based in MEBN, that enables a probabilistic distribution over over models of any FOL axiomatized theory. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 23
  • 46. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes Architecture Version Layers Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 24
  • 47. ©2008 Rommel Novaes Carvalho – University of Brasília UnBBayes Architecture Extension Points Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 25
  • 48. ©2008 Rommel Novaes Carvalho – University of Brasília MEBN before UnBBayes Protégé GUI Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 26
  • 49. ©2008 Rommel Novaes Carvalho – University of Brasília MEBN after UnBBayes UnBBayes GUI Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 27
  • 50. ©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Star Trek Model Starship Zone TimeStep [ord] SensorReport Sensor CloakMode Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 28
  • 51. ©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Resident node L OW P R- L OW PR- Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 29
  • 52. ©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Input node E BN M L OW BN P R- M E E BN M Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 30
  • 53. ©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Context node Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 31
  • 54. ©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Proposed grammar for dynamic CPT E BN M ithf w i w ted noes n Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 32
  • 55. ©2008 Rommel Novaes Carvalho – University of Brasília Building a MTheory Saving ubf and pr-owl files UnBBayes Save UBF PR-OWL Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 33
  • 56. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Inserting entities in the KB Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 34
  • 57. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Inserting evidences in the KB Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 35
  • 58. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 59. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- LIST), search for evidence in the KB. If found it, return it. If not, continue. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 60. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- LIST), search for evidence in the KB. If found it, return it. If not, continue. ii. Search for the resident node that has the name NODE and get its MFrag. Once NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 61. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- LIST), search for evidence in the KB. If found it, return it. If not, continue. ii. Search for the resident node that has the name NODE and get its MFrag. Once NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. iii. Verify in the KB which context node refers to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 62. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. ii. Search for the resident node that has the name NODE and get its MFrag. Once NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. iii. Verify in the KB which context node refers to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 63. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. v. For each father of NODE, go to step (i), ii. Search for the resident node that has the replacing the OVs by the known entities name NODE and get its MFrag. Once (contained in the query or KB). NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. iii. Verify in the KB which context node refers to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 64. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. v. For each father of NODE, go to step (i), ii. Search for the resident node that has the replacing the OVs by the known entities name NODE and get its MFrag. Once (contained in the query or KB). NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. vi. Create the NODE’s CPT. iii. Verify in the KB which context node refers to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 65. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. v. For each father of NODE, go to step (i), ii. Search for the resident node that has the replacing the OVs by the known entities name NODE and get its MFrag. Once (contained in the query or KB). NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. vi. Create the NODE’s CPT. iii. Verify in the KB which context node refers vii. Finish. to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, mark the MFrag to use the default distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 66. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Proposed algorithm to create a SSBN i. For a given entry in the form NODE(ENTITY- iv. If the context node in (iii) does not have a LIST), search for evidence in the KB. If solution, create it as father of NODE. found it, return it. If not, continue. v. For each father of NODE, go to step (i), ii. Search for the resident node that has the replacing the OVs by the known entities name NODE and get its MFrag. Once (contained in the query or KB). NODE(OV-LIST) is found, verify if the type in ENTITY-LIST is the same as OV-LIST. vi. Create the NODE’s CPT. iii. Verify in the KB which context node refers vii. Finish. to the OVs in OV-LIST, replacing its values by ENTITY-LIST. If any of them is false, BN mark the MFrag to use the default ME distribution. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 36
  • 67. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Query HarmPotential(!ST4, !T3) = ? Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 37
  • 68. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK OK OK OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 38
  • 69. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 39
  • 70. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 40
  • 71. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K GE OK O AN 2R OK S ER P HA ] )= [F ,! T0 T4 PT (!S a teC FO er D g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 41
  • 72. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK C PT ate er g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 42
  • 73. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK C PT ate er g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 43
  • 74. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 44
  • 75. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 45
  • 76. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O ] [F Z2 z =! Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 46
  • 77. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 47
  • 78. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK C PT r ate e ne g Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 48
  • 79. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK C PT r ate e ne g Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 49
  • 80. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O ] [F Z2 z =! Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 50
  • 81. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 51
  • 82. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK C PT ate er g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 52
  • 83. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O C PT r ate e ne g Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 53
  • 84. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O C PT ate er g en Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 54
  • 85. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O PT OK a teC n er ge Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 55
  • 86. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Bottom-up algorithm (goal driven)‫‏‬ OK K OK O OK Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 56
  • 87. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes BN inference Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 57
  • 88. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes BN inference Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 58
  • 89. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes BN inference Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 59
  • 90. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Explosive* (goal driven with evidence below)‫‏‬ SSBN topology 1 3 Does NODE has children RNs in the same MFrag? If 1 Evidence for yes, call 1 for each NODE in KB? If CHILD. yes, finish. 4 Get related INs to 2 Get related NODE. If INs have context nodes children, call 1 for and evaluate each CHILD. them. OK OK OK OK 5 If NODE is permanent, evaluate NODE’s parents. Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 60
  • 91. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Explosive* (goal driven with evidence below)‫‏‬ Permanent Nodes 1 1 The query node is 1 4 always permanent. 2 Parents from a 2 3 5 permanent node are also permanent (except for evidence nodes). 3 All findings are permanent nodes. 1 6 4 Parents are only evaluated if node is permanent. 2 3 5 Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 61
  • 92. ©2008 Rommel Novaes Carvalho – University of Brasília SSBN in UnBBayes Explosive* (goal driven with evidence below)‫‏‬ General algorithm 1 Finding No parent 1 Generate SSBN topology. 2 Create CPT for permanent nodes. 3 Remove temporary nodes. 4 Compile the network. 5 Set evidences and update beliefs. Finding below Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 62
  • 93. ©2008 Rommel Novaes Carvalho – University of Brasília General Metaphor Simple use of BN for end users Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 63
  • 94. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
  • 95. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
  • 96. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; 2. Identify the entities present in the domain. If you are having trouble trying to come up with different entities, you probably do not need to use MEBN. Try using just BN, instead; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
  • 97. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; 2. Identify the entities present in the domain. If you are having trouble trying to come up with different entities, you probably do not need to use MEBN. Try using just BN, instead; 3. Identify logical groups (there are groups of information that are or could be logically put together) for identifying sets of entities; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
  • 98. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; 2. Identify the entities present in the domain. If you are having trouble trying to come up with different entities, you probably do not need to use MEBN. Try using just BN, instead; 3. Identify logical groups (there are groups of information that are or could be logically put together) for identifying sets of entities; 4. Identify criterias that can classify in some way the identified entities. This will help you choose which entities are relevant and which are not to solve the problem (discard them from your model). This step can also help to detect uncertain about the existence of information and to identify contexts where certain informations can be considered valid; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
  • 99. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 1. Identify and define the scope of the problem to be solved; 2. Identify the entities present in the domain. If you are having trouble trying to come up with different entities, you probably do not need to use MEBN. Try using just BN, instead; 3. Identify logical groups (there are groups of information that are or could be logically put together) for identifying sets of entities; 4. Identify criterias that can classify in some way the identified entities. This will help you choose which entities are relevant and which are not to solve the problem (discard them from your model). This step can also help to detect uncertain about the existence of information and to identify contexts where certain informations can be considered valid; 5. Identify the attributes the entities can have; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 64
  • 100. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
  • 101. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
  • 102. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
  • 103. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; 8. Evaluate if MEBN is necessary and/or sufficient for modeling the problem; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
  • 104. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; 8. Evaluate if MEBN is necessary and/or sufficient for modeling the problem; 9. Evaluate if the identified entities are enough for your MEBN model; Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
  • 105. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; 8. Evaluate if MEBN is necessary and/or sufficient for modeling the problem; 9. Evaluate if the identified entities are enough for your MEBN model; 10. Map the entities, groups, rules, and relations identified to their respective MEBN element (MFrag, Node, etc); Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
  • 106. ©2008 Rommel Novaes Carvalho – University of Brasília Building MEBN Models A simple guideline 6. If any attribute identified in the previous step is continuous, decide how the discretization must be done. In the case where the values are discrete, but there are too many values, try to group them some how; 7. Identify rules related to entities present in the same group and rules related to entities in different groups; 8. Evaluate if MEBN is necessary and/or sufficient for modeling the problem; 9. Evaluate if the identified entities are enough for your MEBN model; 10. Map the entities, groups, rules, and relations identified to their respective MEBN element (MFrag, Node, etc); 11. Design the model in UnBBayes. Here, you might need to change your model a little bit, because UnBBayes has some singularities due to implementation (the way it implements recursion, possible states for a resident node, etc). Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 65
  • 107. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 66
  • 108. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Results Contributions to PR-OWL/MEBN Tool for representing and reasoning in probabilistic ontologies Complete use case example Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 66
  • 109. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Results Contributions to PR-OWL/MEBN Tool for representing and reasoning in probabilistic ontologies Complete use case example Scientific contributions PR-OWL Entity as a possible state for a node Global exclusivity Built-in recursion MEBN New algorithm for creating SSBN Grammar for creating dynamic CPT Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 66
  • 110. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 67
  • 111. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Technological contributions New algorithm for creating SSBN implemented First implementation of PR-OWL/MEBN in the world GNU GPL v3, open source and free www.sourceforge.net/projects/unbbayes Platform independent – Java Internationalization OWL compatible Friendly GUI Compiler for dynamic CPT Format .ubf Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 67
  • 112. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 68
  • 113. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Limitations Query with only one node No recursion explicit stop condition GUI Node’s size is not proportional to its label Edges do not get to the node’s boundary No overall MTheory visualization No MFrag reuse Some elements present in Laskey 2007 were not implemented Save finding in PR-OWL Built-in recursion in PR-OWL “Likely” and “FOL” findings Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 68
  • 114. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 69
  • 115. ©2008 Rommel Novaes Carvalho – University of Brasília Conclusions Future work Real use case Transparency and corruption prevention CGU - Brazilian General Comptroller Office ONR? Implement limitations Integrate OWL reasoner with MEBN reasoner => real PR-OWL reasoner MCMC (Gibbs) for approximate inference OOBN ...? Motivation - Methodology - Understanding MEBN - Building a MTheory - SSBN in UnBBayes - Conclusions 69
  • 116. ©2008 Rommel Novaes Carvalho – University of Brasília Obrigado! Slides Potpourri Rommel Novaes Carvalho and GIA (Artificial Intelligence Group) from UnB GMU - September 19th 2008