Adaptive Loose Coupling Intelligent Rule System ALCIRS Research Skills – The Presentation Presented by: Irfan Subakti – 1054257 Supervisor: Prof. John A. Barnden School of Computer Science University of Birmingham United Kingdom 25 January 2012 Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 1
Ideas Adaptive Able to adjust to another type of situation Loose coupling Overcome rule dependency changing problem Intelligent Learn for improving its rule semantic understanding, rule learning & generating Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 2
Motivation Variable-Centered Intelligent Rule System (VCIRS) - Subakti (2005, 2006, 2007) Monotonically increasing Rule Based (RB) as time goes by Tight coupling inflexibility in rule changing Too simple rule generating need more creativity Blackboard Systems – Erman et al. (1980), Corkill (1991) Dealing with complex applications which are roughly defined flexible in representation & in contributing problem solving Disadvantage: formal specification Contextual Ontologies – Benslimane et al. (2006) A concept’s set of properties is vary depend on a context Semantic Understanding – Shih et al. (2011) Capturing the interpretation of the behaviours and situations Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 3
Motivation (cont’d) Robust Growing Neural Gas (RGNG) – Kin and Suganthan (2004) Robust properties in clustering Outlier resistance Adaptive modulation of learning rates Cluster repulsion Insensitivity Initialization Input sequence ordering Outlier presence Particle Swarm Optimization (PSO) - Kennedy and Eberhart (2001) Simple idea with outstanding result in optimisation Creativity in Reasoning – Indurkhya (1997) New categories & interpretation can be created in legal reasoning Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 4
Basic Component Blackboard model Knowledge Sources Blackboard (Corkill, 1991) Control Component Rule Based ALCIRS Rules Interface Implement Ontology Inference Engine Contextual Ontology Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 5
Basic Framework PSO (Kennedy and Eberhart, 2001 ) RGNG (Kin and Suganthan, 2004) Rule generating creativity (Indurkhya, 1997) Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 6
Basic Framework (cont’ed) Basic BS Blackboard Executing Library of KS Activation KSs (Corkill, 1991) Events Control Pending Components KS Activations Rule Based Inference Engine Rules PSORGNG - clustering Interface Basic ALCIRS Adaptive Implement Ontology Rule dependency Loose Coupling Rule learning, reasoning, generating Intelligence Contextual Ontology Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 7
Adaptive - Methodology Adaptive Raw data Rules will be clustered in proper place, using PSORGNG Existing rules in Rule Base (RB) Interface & Implement parts will be classified, supported by Contextual Ontology Generating new rules Supported by Contextual Ontology Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 8
Loose Coupling - Methodology Loose coupling Rule dependency changing Each rule has Interface o A part that can be shared to other rules global o Other rules may use a little or none of this part o Flexibility concept applied, since all rules loosely can be connected with this part o As a bridge for contextual ontologies layer Implement o a specific part which dedicated to its rule local Ontology o Linked to contextual ontology further rule learning, reasoning & generating Core ontology the lowest level of contextual ontology can be used as the last resort if higher contextual ontologies failed to do so Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 9
Intelligence - Methodology Intelligence Semantic understanding Understand the meaning of rule given a context supported by contextual ontology Rule learning, reasoning & generating Contextual ontology continually learning to optimise the usefulness of the rules Contextual reasoning supported by contextual ontology gives an inference based on the context Core ontology performing creativity in producing a new rule from the existing rules in RB given a new case Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 10
Loose Coupling Looseness definition When a rule only uses none or little part of other rules loose coupling mechanism Part usages on rules None Explicit: direct assignment. E.g., weight = input_weight Implicit: by using the contextual ontology Little part Using exactly the same term. E.g., Rule #1 uses input_weight in its Interface, while Rule #2 also uses input_weight in its Implement Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 11
Case Study (1) Supermarket goods purchasing Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 12
Case Study (2) Owning a car and a house An example of a rule, which has Interface Implement Ontology Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 13
Case Study (3) Owning a car and a house (continued) Loose coupling rules example #1 A user starts creating a new rule #Car-house owning# defining a relation between owning a car and a house #Vehicle type# defining the types of vehicles Loose coupling No part from #Car-house owning# is used in #Vehicle type# Direct assignment: weight = input_weight at #Vehicle type# Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 14
Case Study (4) Owning a car and a house (continued) Loose coupling rules example #2 Another day, the user willing to add up a rule #House type# defining the types of houses Loose coupling Little part from #Car-house owning# is used in #House type# house_type is used in both rules Little part from #Vehicle type# is used in #House type# vehicle_type is used in both rules Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 15
Case Study (5) Owning a car and a house (continued) Loose coupling rules example #3 Then it turned another day and the user willing to add up a rule #Garage type# Loose coupling Little part from #Vehicle type# is used in #Garage type# wheels is used in both rules No part from #Car-house owning# is used in #Garage type# Direct assignment: wheels = input_wheels at #Vehicle type# Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 16
Case Study (6) Owning a car and a house (continued) A whole rules in RB Loose coupling No part from #Car-house owning# is used in #Vehicle type# Direct assignment: weight = input_weight at #Vehicle type# Little part from #Car-house owning# is used in #House type# house_type is used in both rules Little part from #Vehicle type# is used in #House type# vehicle_type is used in both rules Little part from #Vehicle type# is used in #Garage type# wheels is used in both rules No part from #Car-house owning# is used in #Garage type# Direct assignment: wheels = input_wheels at #Vehicle type# Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 17
Rule Generating Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 18
Anticipated Result Adaptive Rule is clustered in certain group establishing a rule for each formed cluster Interface & Implement is classified in certain category New rule can be generated for a new case creativity Loose coupling Rule changing/updating in a given contextual meaning is easily performed without worry about rule dependency Intelligent Comprehend the meaning of rule given a context Optimise the usefulness of the rules Contextual reasoning Able to perform creativity in a new rule creation Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 19
Conclusion Adaptive Loose Coupling Intelligent Rule System (ALCIRS) A Rule-Based system Use a specific framework which works adaptively & intelligently comparing to Blackboard Systems Treating rule in the loose coupling manner Rule dependency in a given context is automatically preserved Rule updating is easily perform without worry about this dependency Rules are clustered and classified automatically Understanding the contextual meaning of given rule Optimising usefulness of the rules Contextual reasoning Perform creativity in the rule generation Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 20
Future Work Continue reading the literature and comprehend it deeper to suit the proposed system Implementing the framework Using some examples Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 21
Adaptive Loose Coupling Intelligent Rule System ALCIRS Thank you for your attention! Research Skills | ALCIRS | Irfan Subakti (1054257) 25 Jan 2012 22
Recommend
More recommend