Artificial neural network library

 

A key issue in the design of a neural networks simulation library is the design of a proper set of data structures. Powerful data structures should allow the incorporation and implementation of a wide variety of neural network paradigms, whereas the clarity and consistency of the data structures are an essential conditions to overcome delay for novice users. In this chapter a number of such design considerations of neural network data structures is described, followed by an explanation of the implementation in SPRANNLIB.

• Considerations on Network Representation
• Implementation
  • The top level network data structure
  • The unit structure
  • The link data structure
  • The weight structure
  • The unit value structure
• Functions and Algorithms
  • Considerations on functions and algorithms
  • Supporting network functions
    • Low level supporting functions
    • High level supporting functions
• Learning algorithms
  • Feed forward networks
  • Kohonen networks