• A large number of units which are connected to others (connectionism) and the connections have different strengths (weights)
  • Units are arranged in layers
  • Computation happens in parallel


  • Like in a classical GOFAI system, representations are assigned to connectionist networks by the people who build them
  • There are two types of connectionist representation
    1. localist → each unit is assigned a feature as a whole that represents
    2. distributed → state of the network as a whole represents something
      • often claimed to be one of the distinctive features of connectionism
      • connectionist networks is sometimes called parallel distributed processing (PDP)
  • Can be trained to learn and resembles the structure of the brain much more closely than any classical computer