demo:

Each directory contains a set of demonstration projects for the 
various algorithms.  These are useful for exploring the algorithms
and understanding how to accomplish certain things with them.

Basic examples of the various algorithms:

bp 	use the bp++ program to explore XOR, and build your own 424
	encoder. both examples are the subject of the tutorial
	chapter in the manual.

rbp	two versions of recurrent backprop applied to the xor task.
	one is real-time continuous recurrent backprop through time,
	which is capable of learning trajectories over time, and the other
	is the Almeida-Pineda algorithm, which learns to settle into
	an attractor.  both simply settle into stable attractor states.

cs	the constraint satisfaction system is shown with two examples--one
	is yet another different version of the xor problem--this time 
	using a stochastic network that learns to match a distribution of
	different patterns on a set of output units, where the patterns
	are the input-output values of xor.  the other example is a large
	constraint satisfaction network that processes ambiguous displays
	and decides which region is figure and which is ground.  this one 
	shows how to set up complicated pre-wired connectivity patterns.

so	self-organizing learning algorithms!


Misc examples of other software features:

bp_misc examples of various other things to do with bp, and other
	aspects of the software.

bp_srn 	a simple-recurrent network that learns to act like a
	finite-state-automaton.

bridge	shows how to link together two networks, which could be using
	different algorithms, and have processing in one influence
	processing in the other.

css	example code for compiling a hard-coded version of a css file.
	also a useful utility for analysing log files generated by pdp++.
