2. Topic 4: Overview
• Agents & Intelligent Agents
• Agent Models
• Bene
fi
ts of user adaptivity
• Usability challenges
• Collecting data from users
• Future needs in IUI’s
2
4. Characteristics of Agents in
Complex Systems
• Heterogeneity
• Local Interactions
• Adaptation
• Emergence
• Self-organisation
5. Examples of Agents in
Complex Systems
• Ant Colonies
• Economic Markets
• Immune System
• Traf
fi
c Systems
6. Importance of Understanding
and Managing Complex Systems
• Predicting System Behaviour
• Designing Arti
fi
cial Complex Systems
• Managing Natural and Social Systems
7. Agent-Based Models
• Capturing Emergent Phenomena
• Exploring Heterogeneity and Adaptation
• Simulating Non-Linearity and Feedback Loops
• Providing a Natural Description of Systems
• Flexibility and Scalability
8. Environment as a Problem
Space for Agents
• Agent-Environment Interaction
• Control and In
fl
uence
• Non-Determinism
• Effectoric Capabilities
• Preconditions for Actions
9. Environment Properties affecting
Agent Decision-Making
• Accessible vs. Inaccessible
• Deterministic vs. Non-Deterministic
• Episodic vs. Non-Episodic
• Static vs. Dynamic
• Discrete vs. Continuous
10. Task Environments in AI
• Fully Observable vs. Partially Observable
• Deterministic vs. Stochastic
• Competitive vs. Collaborative
• Single agent vs. Multi-agent
11. Task Environments in AI
• Dynamic vs. Static
• Discrete vs. Continuous
• Episodic vs. Sequential
• Known vs. Unknown
12. Agent Interaction
• Perception
• Decision Making
• Action
• Feedback Loops
• Communication (in Multi-agent Systems)
• Learning and Adaptation
14. Intelligent Agent Examples
• Telecommunication Systems
• Personal Digital Assistants
• Information Management
• Information Economies
• Business Applications
• Air Traf
fi
c Control
• Computer Simulation
• Transportation Management
• Financial Management
• Healthcare
• Smart Homes
• Robotics
• Gaming
• Autonomous Vehicles
• Customer Service
15. Interface Agents
• Roles and Components
• Applications:
• AI Assistants
• Machine Learning
• IT Service Management
• Network Management
• Software Development
16. Interface Agents Examples
• Email Sorting Agents
• Personal Digital Assistants
• R e c o m m e n d a t i o n
Systems
• Calendar Management
Agents
• Navigation and Travel
Agents
• Learning Interface Agents
• Character-based Agents
• Microsoft Of
fi
ce Assistant
• Web Browsing Assistant
• Automated Customer
Service Agent
• H e a l t h a n d F i t n e s s
Coaches
• Smart Home Controllers
17. How Interface Agents Learn
about their Environment
• Direct Programming
• Machine Learning
• Reinforcement Learning
• Observation and Imitation
• User Feedback
• Training by Example
• Communication with other Agents
• User and Domain Models
18. How Interface Agents
Perceive their Environment
• Sensors
• Data Processing
• Direct Programming and Rules
• Machine Learning and Adaptation
• Communication with Other Agents
• Active vs. Passive Sensing
• Perception Policies
19. ML Algorithms used by IAs
to Perceive Environment
• Deep Learning
• Reinforcement Learning
• Interactive Machine Learning
• Deep Neural Networks
• Terrain Traversability Analysis
• Vision Deep Learning
20. Unsupervised ML Techniques to
Improve Perception Techniques
• Clustering
• Dimensionality Reduction
• Anomaly Detection
• Association Rule Learning
• Feature Learning
21. Challenges when using
Unsupervised ML Techniques
• Critical Flaws and Prospective Possibilities in Data
• Veri
fi
cation, Validation and Trust
• Absence of Guidance and Nonsensical Results
22. Reinforcement Learning for
Interface Agents
• Agent-Interface Environment in Reinforcement
Learning
• Reinforcement Learning Agents
• Learning Interface Agents
• Challenges and Considerations
23. Reinforcement Learning
Algorithms for Interface Agents
• Deep Q-Network
• Proximal Policy Optimisation
• Deep Deterministic Policy Gradient
• Soft Actor Critic
• Rainbow
• Monte Carlo Tree Search
• Reinforce
24. Challenges for using Reinforcement
Learning for Interface Agents
• Sparse and Delayed
Rewards
• Exploration vs. Exploitation
Dilemma
• Sample Ef
fi
ciency
• High Dimensional State
Space
• C r e d i t A s s i g n m e n t
Problem
• Generalisation Across
Tasks
• S a f e t y a n d E t h i c a l
Concerns
• Partial Observability
• Scalability
• Evaluation and Validation