Title: Local adaptive coordinate systems, an alternative to neural algorithms

  • Objective: formulate an alternative to neural networks based on parametrization of local adaptive systems
  • Participants: W. Duch
  • Time:  since 1994, but low priority
  • Results: some notes and one master thesis; this idea seems to be similar to the general minimal distance method; topographical interpolation in SOM goes in similar direction.
  • Open problems: find the best way to parametrize the input/output space metrics depending on available data; check Hecht-Nielsen results on natural metrics (Iizuka conference). This project moved in the direction of minimal distance methods; see here.

  •  
  • Working Log (local access only)

  •