The command
mrun xor.par
with the following configuration file xor.par
par processor 0 for 4 cntpar -N2 -O1 -p200 -b5 -e0.001 -f1 -l2 networkis ternarytree endpar
starts a parallel simulation of the XOR-problem
(1-norm with -N2 and -O1)
on a ternary tree of 4 transputers, using
the combined genetic backpropagation algorithm with
a population size of 
( -p200) and 5
backpropagation steps per generation ( -b10).
The simulation stops when the average error per pattern
of the best individual is lower than 0.001 ( -e0.001).
Log output is produced every second generation ( -l2) and the resulting network is printed ( -f1):
Simulation Parameters:
Network: 2 bit Counter (counts 1s in input) (4 patterns)
Topology 2-2-1, 3 Neurons, 88 bits (Weights 8)
Weights in [-10.00, 10.00]
5 backpropagation steps per generation
Simulation: Procs = 4, PopSize = 200, MaxGen = 10000, MaxErr = 0.0010
RandomSeed = 802499780
Genetic: Pcopy = 0 %, Pmutate(1..2 pt.) = 60 %, Pcrossover = 40 %
Selection 100 % linear, 0 % uniform
Backprop.: Method: Batch, InitWeight = [ -1.000, 1.000]
Gen 1: t= 0, MinErr= 0.1242, AvgErr= 0.1335
Gen 2: t= 1, MinErr= 0.1225, AvgErr= 0.1330
Gen 4: t= 2, MinErr= 0.1193, AvgErr= 0.1335
Gen 6: t= 3, MinErr= 0.1087, AvgErr= 0.1286
Gen 8: t= 4, MinErr= 0.0890, AvgErr= 0.1252
Gen 10: t= 5, MinErr= 0.0574, AvgErr= 0.1134
Gen 12: t= 6, MinErr= 0.0202, AvgErr= 0.1022
Gen 14: t= 7, MinErr= 0.0090, AvgErr= 0.0684
Gen 16: t= 8, MinErr= 0.0040, AvgErr= 0.0378
Gen 18: t= 9, MinErr= 0.0014, AvgErr= 0.0252
Gen 20: t= 10, MinErr= 0.0012, AvgErr= 0.0195
Gen 21: t= 11, MinErr= 0.0006, AvgErr= 0.0160
Statistic total per sec. time
--------------------------------------------------------
Generations: 21 1.909091 0.523810 s
Individuals global: 4200 381.818176 0.002619 s
Individuals local: 1050 95.454544 0.010476 s
Backprop. steps: 21000 1909.090942 0.000524 s
evaluated Patterns: 25200 2290.909180 0.436508 ms
(-8.75, 7.80,b:-3.18)( 3.88,-3.80,b:-1.53)
( 9.69, 9.14,b:-5.14)
program succeeded.