Bayesian semantics for the semantic web




НазваниеBayesian semantics for the semantic web
страница20/20
Дата06.10.2012
Размер1.34 Mb.
ТипДокументы
1   ...   12   13   14   15   16   17   18   19   20
n tracks we will have to have n Bayesian Networks for assessing its respective nature and potential danger. Figure 43 shows the general schema of the decision process to be performed after all data on tracks is received.



  1. Wise Pilot system – general scheme

In the bottom of the picture we have the n Bayesian Networks related to the n tracks. The system works in discrete time, which means that at time t it will collect all information from the n tracks, use n BNs for assessing its nature and potential danger, and then evaluate what is the best combination of the four decisions it has to take (i.e. what is the best decision policy) for that specific situation given the objectives (i.e. attack target, avoid fratricide, and maximize survivability) and the most updated information available at time t.

The advantages of a Bayesian Inference system over deterministic rule-based systems for dynamic decision situations such as the fighter pilot problem have been discussed extensively in the literature (e.g. Costa, 1999). However, a major obstacle for implementation of this and similar systems is the lack of representational power of Bayesian Networks for dealing with the variable number of tracks for each time t.

As an example, Figure 44 shows how the system would look like in a given time t in which we have four tracks being perceived by the system.



  1. Wise Pilot with 4 Tracks

Now suppose that in time t+1 (i.e. on the next system iteration) two new tracks are perceived by the system and one of the tracks from time t went away. In this new situation, the system depicted in Figure 44 is no longer valid, while the new configuration should look like the one illustrated in Figure 45.

In other words, in a highly dynamic environment like the one covered in this example, a Bayesian Inference system would have to be reconfigured almost for every iteration, greatly increasing the complexity of its implementation. Furthermore, there may be uncertainty about the correct configuration at any given iteration. Even more problematic is the situation in which we are unsure whether a sensor report indicates a real or spurious object, or whether two reports refer to the same or different subjects. In these cases, the number of instances of the track sub-network is uncertain.

Behind this limitation is the fact that Bayesian Networks have limited expressive power, while in situations like the one portrayed (and in many interesting situations of the real world as well) a more powerful representational formalism is desired.

More specifically, Bayesian Networks allow probability statements (i.e. propositions) over specific instances of a model, but do not support making general assertions over non-specific instances (e.g. statements about variables instead of unique instances). Thus, when facing problems like the fighter pilot’s we need a language that combines the inferential power of Bayesian Networks with the representational power of first-order logic.



  1. Wise Pilot with 5 Tracks

We have seen in Chapter 3 that MEBN logic provides such combination of Bayesian plausible reasoning and first-order logic expressiveness, and thus is a perfect match for the requirements. This is no surprise, since the suitability of MEBN logic for C3I decision systems was already pointed out in some recent research work (Costa et al., 2005). However, for the very same reasons of the DTB project cited earlier in this appendix, MEBN alone would not guarantee that such system would be interoperable, easily maintainable and upgradeable.

Indeed, in order to realize the concept of Network Centric Warfare being sought by most modern armed forces (cf. Alberts et al., 1999), massive investments must be made to achieve sensor interoperability and information sharing between combatant platforms in a tactical environment.

In the Wise Pilot case, building the system’s multiple ontologies (e.g. ontologies on ground-based radar systems, airborne radars and respective platforms, interceptor aircraft and respective weapon systems, etc.) using PR-OWL would bring the intrinsic advantages of probabilistic ontologies, such as built-in ability to learn from previous engagements, and the possibility of improving maintainability, interoperability (among the system’s ontologies and exterior ones as well), and expandability.

Of course, the above-cited advantages can be easily transposed to similar data fusion systems in the military domain and in civilian applications as well, clearly exposing the promising aspect of the technology outside the Semantic Web framework.

Curriculum Vitae

Paulo Cesar G. da Costa was born in Rio de Janeiro on March 3, 1965, and is a Brazilian citizen. He graduated with honors in the Brazilian Air Force Academy in 1986 and started his career as a fighter pilot, having flown nearly 1.800 hours in fighter aircraft such as the Brazilian-Italian made AM-X. During this period, he pursued specialization in the Electronic Warfare field, attending courses in both Brazil and England, acting as Electronic Warfare Officer in the Brazilian Air Force’s first AM-X Squadron, and then as an invited lecturer in most of the courses from the BAF’s Electronic Warfare Center (CGEGAR). In 1995, he graduated first place out of 115 students, all in the rank of Captain, in the BAF’s EAOAr, a major career course. In 1997, he moved to the USA and started his Master of Science degree in Systems Engineering at George Mason University, where he graduated in 1999 with GPA 4.0 and received the GMU’s Academic Excellence Award. He also received the C3I Certificate from GMU in 1999. Back to Brazil in 2000, he served in the Air Force Chiefs of Staff, where he participated in many projects in the IT field. In 2003, he graduated with honors from the BAF’s ECEMAR, another major career course, and returned to the USA to pursue his PhD degree in Information Technology at George Mason University.



1 ISO is actually a word that was derived from the Greek isos, meaning "equal".

2 Dating established by John Darnell, in his 1990s studies of rock carvings at Wadi el-Holi made by Semitic workers within the Egyptian society. For more information on alphabets and its origins see http://www.xasa.com/wiki/en/wikipedia/a/al/alphabet.html (as accessed in Sept 02, 2004).

3 See table at page 47 for a direct comparison among data mining algorithms.

4 From the W3C Semantic Web page, http://www.w3c.org/2001/sw/, as extracted in June 16, 2005.

5 A markup language adds computer-understandable codes (markups) to convey metadata information within a text file. Depending on the language used, this metadata can be mostly restricted to styling and layout (e.g. HTML) or also include semantic information and other advanced features (e.g. OWL).

6 Emphasis added.

7 The term metaphysics means beyond the study of physics

8 Some degree of cardinality exists in XML Schema

9 Available at http://www.defenselink.mil/nii/bpr/bprcd/0039.htm, as of July 6, 2005.

10 MYCIN was an expert system developed in the seventies to assist medical specialists in the diagnosis of infectious blood diseases, having achieved a performance comparable with that of human experts.

11 DENTRAL was also an expert system in the area of mass spectrometry. Even though its events were inherently probabilistic, this was ignored by the inference engine in favor of a simpler, binary decisions about occurrence or non occurrence of those events

12 Star Trek and related marks are registered trademarks of Paramount Pictures.

13 In spite of both being called languages, markup languages are very different from programming languages. They are static and do not process information, but only store it in a structured way.

14 HTML has a strong focus on displaying information. Even its limited, implied semantics are largely ignored. As an example, tags h1, h2, …, h5 are commonly employed as a formatting tool, rather than to identify header levels in a document structure.

15 The interested reader will find further information on DAML at http://www.daml.org/ and on OIL at http://www.ontoknowledge.org/oil/

16 Bayesian network screen shots were constructed using Netica, http://www.norsys.com.

17 The interest reader can find further information on the Star Trek series in a plethora of websites dedicated to preserve or to extend the history of series, such as www.startrek.com, www.ex-astris-scientia.org, or techspecs.acalltoduty.com.

18 Please, note that standard MEBN logic does not support polymorphism. However, an extension to a typed polymorphic version is proposed in Chapter 4, and would permit a random variable to be resident in more than one MFrag.

19 State names in this Dissertation are alphanumeric strings beginning with a letter, including True and False. However, Laskey (2005) uses the symbols T for True, F for False, and  for Absurd, and requires other state names to begin with an exclamation point (because they are unique identifiers)

20 For efficiency reasons, most knowledge-based model construction systems would not explicitly represent root evidence nodes such as Cloak­Mode(!ST0) or !T1=!T0 or barren nodes such as ZoneFShips(!Z0) and ZoneFShips(!Z0). For expository purposes, the approach taken here was the logically equivalent, although less computationally efficient, approach of including all these nodes explicitly.

21 The alert reader may notice that root evidence nodes and barren nodes that were included in the constructed network of Figure 8 are not included here. As noted above, explicitly representing these nodes is not necessary.

22 Absorption changes the structure of the already-observed length MFrags by removing their dependence on the global average length and setting their observed values to probability 1. It also removes the finding MFrags for these random variables.

23 In classical logic, the terms predicate and function are used in place of Boolean and non-Boolean random variables, respectively. Predicates must have value True or False, and cannot have value Absurd.

24 Even thought the MTheory in Figure 12 was built using the standard version of MEBN, for simplicity and in order to be facilitate its translation to Quiddity*Suite we implicitly allowed polymorphism in the specific case of the IsA MFrags (i.e. by making them binary predicates in that model).

25 Refer to the examples and discussions in the end of subsections 1.8.1 and 1.19.2

26 Available for download at http://protege.stanford.edu/

27 Available for download at http://www.mindswap.org/2004/SWOOP/

28 See http://webode.dia.fi.upm.es/WebODEWeb/index.html

29 Numbers inside brackets refer to the equally numbered circle labels in the pictures

30 ARDA – Advanced Research and Development Activity (www.ic-arda.org)

31 IET – Information Extraction and Transport, Inc. (www.iet.com)

32 AAA is the acronym for Anti-Aircraft Artillery, which includes all gun-based weapons employed against airborne targets (aircraft, helicopters, cruise missiles, etc).
1   ...   12   13   14   15   16   17   18   19   20

Похожие:

Bayesian semantics for the semantic web iconQueries: Enabling Querying for Semantic Associations on the Semantic Web

Bayesian semantics for the semantic web iconSemantic Web : a guide to the Future of xml, Web Services, and Knowledge Management

Bayesian semantics for the semantic web iconThe Semantic Model contains a high level description of the Actions that operate on the objects and attributes in the model. This document does not describe the mapping of the semantics onto a specific protocol or network environment

Bayesian semantics for the semantic web iconЗадача: Предоставляет пользователю унифицированную среду содержащую сервисы и функциональности, варьирующиеся от информационных сервисов и сервисов извлечения
По аналогии с Semantic Web, Semantic Grid может быть определен как расширение современных grid, где информация и сервисы имеют четкий...
Bayesian semantics for the semantic web iconThis document contains a list of references to publications and reports about Bayesian Net technology, and especially Bayesian Net applications. The report will

Bayesian semantics for the semantic web iconValentina Janev, Sanja Vranes, Semantic Web Tools and Technologies for Competence Management, Lambert Academic Publishing GmbH, isbn: 978-8454-4166-5, Sarbrucken, Germany, 2011

Bayesian semantics for the semantic web iconИнтеллектуальное реферирование: онтологический подход и его реализация в решениях Ontos
Обсуждению вопросов создания системы реферирования под управлением онтологий, разработанной в рамках решений Ontos для Semantic Web,...
Bayesian semantics for the semantic web iconWeb dizains web design
Основные компоненты web-страницы и способы их визуального представления на страницах сайта
Bayesian semantics for the semantic web iconDependency Tree Semantics and Underspecification

Bayesian semantics for the semantic web iconSemantics of contour lines' spatial relation

Разместите кнопку на своём сайте:
Библиотека


База данных защищена авторским правом ©lib.znate.ru 2014
обратиться к администрации
Библиотека
Главная страница