Download presentation
Presentation is loading. Please wait.
Published byAnis Perry Modified over 8 years ago
1
Machine Understanding in Agent-Directed Simulation: State-of-the-Art and Research Directions Tuncer Ören University of Ottawa School of EE and Computer Science Ottawa, ON, Canada, K1N 6N5 oren@eecs.uOttawa.ca oren@eecs.uOttawa.ca Levent Yilmaz Dept. of Computer Science and Software Engineering Auburn University Auburn, AL, USA yilmaz@auburn.edu Nasser Ghasem-Aghaee Department of Computer Eng. Sheikh Bahaee University & University of Isfahan, Isfahan, Iran aghaee@eng.ui.ac.ir Mohammad Kazemifard Department of Computer Eng. Razi University Kermanshah, Iran m.kazemifard@razi.ac.ir Fariba Noori Department of Computer Eng. Razi University Kermanshah, Iran fa.noori@pgs.razi.ac.ir 2016 Spring Simulation Multi-Conference (SpringSim'16) April 3 - 6, 2016, Pasadena, CA, USA 2016 Modeling and Simulation of Complexity in Intelligent, Adaptive and Autonomous Systems (MSCIAAS) Symposium http://www.site.uottawa.ca/~oren/y/2016/D3_u-pres
2
- Definitions - Our research - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types - in 4 groups - Four categories of machine understanding - Emotions - Emotion understanding - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation Plan
3
Wide scope of the meanings attached to understanding: to possess a passive knowledge of (a language) to seize the meaning of to grasp the reasonableness of to have thorough or technical acquaintance with or expertness in the practice of to be thoroughly familiar with the character to accept as a fact or truth or regard as plausible without utter certainty to believe or infer something to be the case to show a sympathetic or tolerant attitude toward to interpret in one of a number of possible ways to form a reasoned judgment concerning (something) to have the power of seizing meanings, forming reasoned judgments...
4
“Thus, if a system knows about X, a class of objects or relations on objects, it is able to use an (internal) representation of the class in at least the following ways: -receive information about the class, -generate elements in the class, -recognize members of the class and -discriminate them from other class members, -answer questions about the class, and -take into account information about changes in the class members.” An early (1986) clarification of the relationship of understanding and knowledge was given by B.P. Zeigler: Zeigler, B.P. (1986). System Knowledge: A definition and its implications. In: M.S. Elzas, T.I. Ören, and B.P. Zeigler (eds.) (1986). Modelling and Simulation Methodology in the Artificial Intelligence Era. North-Holland, Amsterdam, pp. 15-17.
5
Our definition of (machine) understanding: “Understanding an entity (a thing, a concept, an event, or a system) is a mapping between - the perceived knowledge about the entity and - a meta-model (i.e., a more general knowledge) of the entity. Hence, a machine understanding system, like a human, cannot understand an entity, if it does not have adequate background knowledge about it. Our approach can thus be conceived as “knowledge-driven approach” for machine understanding.
6
- Definitions - Our research: Phases, references - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types - in 4 groups - Four categories of machine understanding - Emotions - Emotion understanding - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation Plan
7
Major stages of our research: 1990-1995 Understanding simulation software; program understanding 1997-2006 Understanding systems 2006-2008 Systems with understanding ability and understanding agents 2009- Agents with ability to understand emotions; switchable understanding 2011- Avoidance of misunderstanding 2015- Enriching machine understanding paradigm 2016- Exploring the synergy of machine understanding and all three aspects of agent-directed simulation (ADS)
8
Ören, T.I., M. Kazemifard, and F. Noori (2015-Feature article). Agents with four categories of understanding abilities and their role in two-stage (deep) emotional intelligence simulation. International Journal of Modeling, Simulation, and Scientific Computing (IJMSSC), vol. 6, issue 3 (September), pp: 1-16.Agents with four categories of understanding abilities and their role in two-stage (deep) emotional intelligence simulation Kazemifard, M., Ghasem-Aghaee, N., Koenig, B.L. and T.I. Ören (2014). An Emotion Understanding Framework for Intelligent Agents based on Episodic and Semantic Memories. Journal of Autonomous Agents and Multi-Agent Systems. January 2014, Volume 28, Issue 1, pp 126-153. Ören, T.I., Ghasem-Aghaee, N., and L. Yilmaz (2007). An Ontology-Based Dictionary of Understanding as a Basis for Software Agents with Understanding Abilities. Proceedings of the Spring Simulation Page 4 of 5 Multiconference (SpringSim’07). Norfolk, VA, March 25-29, 2007, pp. 19-27. (presentation)An Ontology-Based Dictionary of Understanding as a Basis for Software Agents with Understanding Abilitiespresentation ~20 articles and ~20 presentations~20 articles and ~20 presentations (without articles) In this presentation, quotations are used especially from the following:
9
- Definitions - Our research: Phases, references - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types - in 4 groups - Four categories of machine understanding - Emotions - Emotion understanding - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation Plan
10
Elements of a machine understanding system
11
Functional decomposition of machine understanding systems
12
- Definitions - Our research: Phases, references - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types - in 4 groups - Four categories of machine understanding - Emotions - Emotion understanding - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation Plan
13
~ 60 types of machine understanding
14
~ 60 types of understanding in four groups: -Product of understanding -Process of understanding -Meta-knowledge of understanding -Understanding system Understanding: - is a process (understanding process) -of a system (understanding system) - generates a product (product of understanding) - requires a meta-knowledge (meta-knowledge of understanding)
23
- Definitions - Our research: Phases, references - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types - in 4 groups - Four categories of machine understanding - Emotions - Emotion understanding - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation Plan
24
Four categories of understanding: - Basic understanding - Rich understanding - Exploratory understanding - Theory-based understanding
25
A typical rich-understanding system
26
Exploratory understanding - Exploratory understanding is similar to basic understanding or rich understanding. - Except the understanding process starts with a perception (2). - Background knowledge (or a meta-model) (1) has to be found or developed to process the perception.
27
Theory-based understanding - Theory-based understanding starts with a hypothesis (or theory); - Then the necessary technology would be developed to perceive (detect) relevant phenomena. - The hypothesis (or theory) would be tested later. - Examples: gravitational waves; elementary particles.
28
Two-stage (deep) basic understanding Understanding: - one-stage understanding - two-stage understanding
29
Functional decomposition of a two-stage (or deep) understanding system Functional decomposition of a one-stage (or deep) understanding system In two-stage (or deep) understanding system, in addition to a descriptive understanding, an evaluative understanding can also be generated.
30
- Definitions - Our research: Phases, references - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types - in 4 groups - Four categories of machine understanding - Emotions - Emotion understanding - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation Plan
31
Admiration Anger Disappointment Distress Fear Fear-confirmed Gratitude Hope Joy Relief Reproach Satisfaction Basic emotions:
32
Event-based One step Two steps Action-basedOne step Compound
33
Event-based emotions – One step
34
Event-based emotions – Two steps
35
Action-based emotions – One step Standards of the agent Action done by an agent is ConsistentInconsistent AdmirationReproach Compound emotions Joy + AdmirationGratitude Distress + ReproachAnger
36
- Definitions - Our research: Phases, references - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types - in 4 groups - Four categories of machine understanding - Emotions - Emotion understanding - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation Plan
37
“Emotion understanding is a cognitive activity of making inferences using knowledge about emotions about - why an agent is an emotional state (e.g., unfair treatment makes an individual angry) and - which actions are associated with the emotional state (e.g., an angry individual attacks others.” An understanding system can generate: -A one-stage descriptive understanding of an emotion of an agent -A two-stage descriptive as well as evaluative understanding … -A single vision understanding -Multi-understanding and switchable understanding -…
38
Functional decomposition of the framework for agents with one-stage basic understanding abilities for emotional intelligence simulation
39
Functional decomposition of the framework for agents with two-stage basic understanding abilities for emotional intelligence simulation
40
- Definitions - Our research: Phases, references - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types - in 4 groups - Four categories of machine understanding - Emotions - Emotion understanding - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation Plan
41
Agent-Directed Simulation (synergy of simulation and agents) Contributions of Simulation for agents: - Agent simulation (simulation of agent systems) Agents for simulation: - Agent-supported simulation - Agent-monitored simulation It is worth to consider all aspects of simulation such as: -Ören, T.I. (2010). Simulation and Reality: The Big Picture. (Invited paper for the inaugural issue) International Journal of Modeling, Simulation, and Scientific Computing (IJMSSC) (of the Chinese Association for System Simulation - CASS) by the World Scientific Publishing Co. China, Vol. 1, No. 1, 1-25.Simulation and Reality: The Big Picture - 2011 September 7-15, Beijing, China Beijing University of Aeronautics and Astronautics; (aka) Beihang University School of Automation Science & Electrical Engineering Seminar: Modeling and Simulation: Big PictureModeling and Simulation: Big Picture
42
Contributions of Simulation for agents: - Agent simulation Agent-based simulation (simulation of agent systems) Agents for simulation: -Agent-supported simulation -Agent-monitored simulation Agent simulation (for experimentation) is used in many diverse types of applications such as: engineering, management/economy, social systems, psychology, physiology, environmental systems, and military issues. Agent simulation is used to gain experience to develop/enhance three types of skills, namely motor, decision making, or operational skills. Agent simulation is also used to gain experience for entertainment purposes Machine understanding in all three categories of agent simulation
43
Agent-Directed Simulation (synergy of simulation and agents) Contributions of Simulation for agents: - Agent simulation Agent-based simulation (simulation of agent systems) Agents for simulation: - Agent-supported simulation - Agent-monitored simulation Agent-supported simulation: - Use of agents in front-end user interfaces - Use of agents in back-end user interfaces - To process elements of a simulation study symbolically - To provide cognitive abilities to the elements of a simulation study Agent-monitored simulation: - Use of agents during simulation run-time
44
Agent-supported simulation In front-end interfaces: -“Front-end interfaces are used to specify, edit, or generate elements of a simulation problem such as goal of the study, parametric model, model parameters, design of experiments, and experimental conditions for every experiment.” In back-end interfaces: -“Back-end interfaces are used by systems to communicate to the users the primary and auxiliary outputs of the system.” -An early implementation for behavior instrumentation. To process elements of a simulation study symbolically: -Consistency checking To provide cognitive abilities to the elements of a simulation study:
45
Agent-monitored simulation -Agent-monitored simulation is the run-time use of agent technology to monitor and generate model behavior. - Machine understanding provides a basis for context- awareness which leads to context-aware systems that may be very useful in monitoring simulation runs. - An application may be monitoring transitions between several submodels of a multi-model. - Another advanced application area of machine understanding to agent-monitored simulation is monitoring multi- simulation studies.
46
- Definitions: Everyday usage, B.P. Zeigler, our definition - Our research: Phases, references - Machine understanding: Elements, Functional decomposition - Ontological dictionary: ~ 60 types, in 4 groups (product, process, meta-knowledge, system) - Four categories of machine understanding (basic, rich, exploratory, theory-based) - Emotions: event-based, action-based, compound - Emotion understanding: one-stage, two-stage - Machine understanding in agent-directed simulation (ADS) - in agent simulation - in agent-supported simulation - in agent-monitored simulation We have seen – highlights of:
47
Complexity may depend on our point of view. Thank you for your attention!
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.