connectionism is a way of modeling the brain. a connectionist
network consists of nodes and one-way connections.
each node is a simple processing unit (dumber than
a neuron). each node only performs some simple computation.
in the case of most modern connectionist models, they do
some algebra. they do a weighted algebraic sum of all of
their inputs, generating output from some function, which
is generally monotonic but nonlinear. the nodes are not
particularly interesting. the usefulness and interest of a
connectionist network comes from its overall layout. this
is a pattern of connections between the nodes. usually, the
architecture is fixed. nodes are often formed into layers.
each node in a layer is connected to all of the other nodes in its layer,
and all of the nodes in each adjascent layer; in a one- or
two-layer model, all of the nodes are connected to each other.
two-layer models are very common. in these, one layer gets
input of some sort, and the other sends output.
next:
connectionism: node activation