Knowledge representation

Human brain - Long-term memory analogy
1. Declarative Memory
  1. Associated with knowing "what things are" i.e., facts
2. Procedural Memory
  1. Associated with knowing "how to do things" i.e., Methods
Implicit Knowledge
1. Neural networks
  1. Used in powerful pattern matchers
  2. Agent trains and learns in new situations
    1. Sensitive to training set
    2. Implicit formation of knowledge by activation
2. Behavioural based
  1. Rodney Brooks
    1. Post-experience planning
  2. Emergent behaviour
    1. Simple agents produce a non-explicitly programmed complex behaviour
    2. Analogies
      1. Birds swarming
      2. Ants co-operating
3. Chunking
  1. Problem decomposition
    1. Solve & cache solutions to sub-problems
      1. Combine cached solutions forming one solution
Explicit Knowledge
1. GOFAI
  1. Knowledge of World
    1. Observed, acquired, interrogated & revealed
    2. Engineered, reasoned, used & deducted
    3. Hierarchical planning
      1. Meta-level: scheduling, planning, ordering
  2. Lacks & constraints
    1. Reactive planning & problem solving
    2. Environmental needs
      1. Maximum observability
        All aspects relevant to choice of action observable
      2. Deterministic environment
        Saves worrying about uncertainty
      3. Static environment
        Environment doesn't change whilst agent delibrates
    3. Adaptability
2. Facts & Rules
3. Data Structures
  1. Semantic nets
    1. Symbolic representation of relationships between concepts
  2. Frames
    1. Hierarchical structure
      1. Classes, sub-classes
      2. Instances (e.g., name)
      3. Slot:fillers(i.e., description of attributes)

Next up

Knowledge representation

Description

Resource summary

Similar

	Created by Mohamed Moalin over 9 years ago