Simple neural network model for image processing and object recognition (partially complete).
C C++ Shell
Switch branches/tags
Nothing to show
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
src
test
README

README

Basic learning principles:
- opportunity for a neuron to dominate(abstract):
  1). already established large connectivity
  2). frequent firing to acquire connectivity
  3). neurons can form a cluster that always fire together

- opportunity to reconnect:
  1). synapse weakens over time

- neuron potential/energy
  1). collect energy from links, based on link strength;
  2). excite state if energy is larger than threshold;
  3). available energy to distribute is based on state;
  4). distribute available energy to all links
  5). initial energy source : 
      - inputs firing : strong
      - random firing : weak

- competition for new links
  1). each excited neuron connects to the ones 
      with highest energy from previous firing;
  2). then collect/distribute new energy;
  3). cool down after new connection of the next round

- how to suppress duplicated firing?
  If a pattern is already recognized, then do not allow
  new neuron to repeat the same connectivity. Use a look-up table
  of firing signatures; if same signature exist, higher energy
  neuron survive.