Neural Network & Fuzzy Systems

Artificial Intelligence Systems, Fuzzy Systems, And Neural Networks Overlap And Complement One Another

Each of the paradigms has a contribution to make to the ultimate solution of a problem, as is pointed out below:

Symbolic AI systems can contribute with:

  • Rigid theory for symbol manipulation; theorem proving
  • Rigorous exact reasoning mechanisms, including chain reasoning and different reasoning strategies
  • Universal computational models (e.g., the production systems)

Fuzzy systems can contribute with:

  •  Well-developed fuzzy logic theory
  • Humanlike reasoning mechanisms; using linguistic terms
  • Accommodating commonsense knowledge, ambiguous knowledge, imprecise but rational knowledge
  • Universal approximation techniques
  • Robustness, fault tolerance
  • Low cost of development and maintenance
  • Low computational cost

Neural networks can contribute with:

  • Learning from data
  • Modeling empirical behavior of humans
  • Universal approximation techniques
  • Good generalization
  • Methods for extracting knowledge from data
  • Methods for data analysis
  • Associative memories and pattern-matching techniques
  • Massive parallelism
  • Robustness

These three groups of methods as complementary for three reasons:

 (1) They facilitatedealing with different kinds of knowledge representation, different inference, different accuracy andfault tolerance, etc.

 (2) Each of them may be superior to the others when solving a concrete subtask of aproblem; for example, the symbolic approach and the connectionist approach have demonstratedstrengths in solving different tasks. The former has been successful in high-levelcognitive tasks, whereas the latter is more successful in low-level perceptual and learning tasks.

 (3) One method may beapplied to improve the performance of another.