cbp.1

Man entry for cbp (unformatted).


The ci command interpreter is used for the user interface; ? or help
should tell you how it works.  "*" will list the available commands,
and "*=" will list the settings of the various variables.

Most of the variables are irrelevant to standard backpropagation mode.
All the variables with "B" in them, like "display_B", are irrelevent
under any circumstances most users would come across.

The following is a sample session of training a the 424 encoder
example.

	% cd /afs/cs.cmu.edu/user/bap/src/cbp/424
	% ls
	424.env		424.net		424.state
	424.lay		424.slay	424.wet

Crank up the simulator; the network description file, 424.net, is read
at startup; that's what the command line argument is for

	% cbp 424
	11 units, 5 inputs, 4 outputs, 22 connections, 0 equivalences.
	Randomizing Weights, excursion 1.000000, seed 658787246.

Read in the environment file, 424.env.  The "Timesteps" isn't relevent
to standard backprop mode.

	cbp> read_env 424
	Patterns: 4.  Timesteps: 1.

Print values of all variables.  I am placing a tad of documentation
before each one relevent to simple backprop mode.

	cbp> *=
This is printed on the left of each epoch report line, and is used as
the default basename of output files.
	run_name       	
The number of learning cycles that have been done
	epoch          	0
learning rate
	epsilon        	0
	T_epsilon      	0
learning momentum
	momentum       	0
A term $WeightDecay \sum_i w_i^2$ is added to the error
	weight_decay   	0
	T_decay        	0
	end_of_time    	0
	delta_t        	0.1
	forward_criterion 	0.001
	backward_criterion 	1e-06
What kind of error measure to use.  Simple square is just squared
difference of output from target.  Marginated square is zero if output
is within square_margin of target (see below) and squared difference
from there on.
	error_type     	simple square
Relevent only with Marginated square error; see error_type above.
	square_margin  	0.2
Stop the simulation if the angle between the accumulated velocity and
the gradient is greater than 90 degrees.
	stop_slosh     	yes
	stop_high_B    	no
	display_time   	no
	display_B      	no
	B_subepoch_count 	0
	max_B          	0.1
	dB_hack        	1
The kind of network to simulate.
	net_type       	standard backpropagation
Each epoch, show the number of incorrect cases (in parens after
error).  See "error_margin" for criterion.
	display_cases  	yes
Stop the simulation when there are no incorrect cases (as shown by
display_cases, which must be on for this to be enabled.)
	stop_zero_errors 	no
Each epoch, show the ratio between the actual accumulated velocity and
the velocity you'd expect to have accumulated if you reached terminal
velocity, and the gradient was always the same as it is now, and in
the same direction.
	display_expected_velocity 	yes
How near to the target an output must be to count as correct in the
count of correct cases made by activating the display_cases option
	correct_margin 	0.4
Every weight_check_interval, a weight checkpoint file will be written
	weight_check_interval 	0
Weight checkpoint files have this name, with the epoch number tacked
onto the end.  Defaults to run_name.wet
	weight_check_file 	
Every weight_check_interval, a state checkpoint file will be written
	state_check_interval 	0
State checkpoint files have this name, with the epoch number tacked
onto the end.  Defaults to run_name.state
	state_check_file 	
	traj_check_interval 	0
	traj_check_file 	
These determine which states are dumped into state output files.  The
units with numbers betweenn those given, inclusive, are output.  These
default to numbers corresponding to "all the units."
	low_printing_unit       0
	high_printing_unit      10

	perturb_interval 	0
	perturb_magnitude 	0

Get a list of all commands
	cbp> *

This reads the weights from a weight file
	read_weights
This reads a set of I/O patterns 
	read_environment
	read_z0s
	write_z0s
Write the weights out to a named file
	write_weights
Write the states of all the units for each pattern in the environment
out to a named file
	write_states
	write_trajectory
Write the weights to a checkpoint file (see weight_check_file for filename)
	checkpoint_weights
Write the states to a checkpoint file (see weight_check_file for filename)
	checkpoint_states
	checkpoint_trajectory
Run the network for a number of epochs (can be interrupted)
	go
Randomize the weights to uniform random values between plus and minus
x, where x is the "excursion" you give.  A random seed can be given
also, to get the same random weights you got some other time, assuming you
saved the random seed that was used.
	randomize_weights
	one_Ts
Set the accumulated velocity to zero
	clear_momentum
Multiply the learning rate by the given number
	multiply_epsilon
Divide the learning rate by the given number
	divide_epsilon
Change working directories
	cd
Print the current working directory
	pwd
Print a blurb about the ci command interpreter
	help

Set epsilon
	cbp> eps=
	epsilon  (-1e+20 to 1e+20)  [0]  .01
	epsilon        	0.01
Turn on the displaying a count of environmental patterns the net gets wrong
	cbp> display_cases = yes
	display_cases  	yes
Set momentum
	cbp> mom = .8
	momentum       	0.8
Run for a bit
	cbp> go 10
	 1 0.588841(4) cos:NaN |grad|:0.645762 |v|:0.000000 v/E(v):0.000000.
	 2 0.587799(4) cos:0.999999 |grad|:0.644989 |v|:0.645762 v/E(v):0.200240.
	 3 0.585928(4) cos:0.999993 |grad|:0.643583 |v|:1.161598 v/E(v):0.360979.
	 4 0.583402(4) cos:0.999981 |grad|:0.641644 |v|:1.572859 v/E(v):0.490259.
	 5 0.580360(4) cos:0.999961 |grad|:0.639250 |v|:1.899923 v/E(v):0.594423.
	 6 0.576917(4) cos:0.999933 |grad|:0.636459 |v|:2.159171 v/E(v):0.678495.
	 7 0.573165(4) cos:0.999898 |grad|:0.633319 |v|:2.363764 v/E(v):0.746469.
	 8 0.569180(4) cos:0.999858 |grad|:0.629866 |v|:2.524282 v/E(v):0.801530.
	 9 0.565021(4) cos:0.999815 |grad|:0.626132 |v|:2.649223 v/E(v):0.846219.
	 10 0.560738(4) cos:0.999770 |grad|:0.622142 |v|:2.745421 v/E(v):0.882571.
Raise learning rate
	cbp> mul 5
	cbp> go 10
	 11 0.556370(4) cos:0.999726 |grad|:0.617917 |v|:2.818367 v/E(v):0.912216.
	 12 0.534612(4) cos:0.998665 |grad|:0.594130 |v|:2.872478 v/E(v):0.966953.
	 13 0.513663(4) cos:0.997325 |grad|:0.566373 |v|:2.891482 v/E(v):1.021052.
	 14 0.493893(4) cos:0.996029 |grad|:0.535018 |v|:2.878341 v/E(v):1.075978.
	 15 0.475595(4) cos:0.994937 |grad|:0.500645 |v|:2.835967 v/E(v):1.132926.
	 16 0.458977(4) cos:0.994091 |grad|:0.464033 |v|:2.767341 v/E(v):1.192736.
	 17 0.444163(4) cos:0.993444 |grad|:0.426125 |v|:2.675638 v/E(v):1.255800.
	 18 0.431187(4) cos:0.992876 |grad|:0.387951 |v|:2.564304 v/E(v):1.321974.
	 19 0.420007(4) cos:0.992213 |grad|:0.350529 |v|:2.437069 v/E(v):1.390511.
	 20 0.410519(4) cos:0.991242 |grad|:0.314768 |v|:2.297869 v/E(v):1.460038.
Raise it some more
	cbp> mul 5
	cbp> go 10
	 21 0.402571(4) cos:0.989726 |grad|:0.281398 |v|:2.150708 v/E(v):1.528588.
	 22 0.377270(4) cos:0.907307 |grad|:0.142665 |v|:1.999479 v/E(v):2.803034.
	 23 0.368786(4) cos:0.375044 |grad|:0.082195 |v|:1.730064 v/E(v):4.209672.
	 24 0.368351(4) cos:-0.236449 |grad|:0.090744 |v|:1.416928 v/E(v):3.122926.
	Sloshing...
	  Break!
Oops, we raised it too high.  Lower it, clear the accumulated
momentum, and continue
	cbp> div 5
	cbp> clear_mom
	cbp> go 10
	 25 0.368351(4) cos:NaN |grad|:0.090744 |v|:0.000000 v/E(v):0.000000.
	 26 0.368248(4) cos:0.999994 |grad|:0.090641 |v|:0.090744 v/E(v):0.200228.
	 27 0.368063(4) cos:0.999969 |grad|:0.090457 |v|:0.163235 v/E(v):0.360912.
	 28 0.367813(4) cos:0.999912 |grad|:0.090213 |v|:0.221043 v/E(v):0.490050.
	 29 0.367513(4) cos:0.999814 |grad|:0.089925 |v|:0.267042 v/E(v):0.593925.
	 30 0.367172(4) cos:0.999672 |grad|:0.089607 |v|:0.303546 v/E(v):0.677505.
	 31 0.366801(4) cos:0.999488 |grad|:0.089273 |v|:0.332423 v/E(v):0.744733.
	 32 0.366405(4) cos:0.999268 |grad|:0.088933 |v|:0.355177 v/E(v):0.798755.
	 33 0.365992(4) cos:0.999019 |grad|:0.088595 |v|:0.373025 v/E(v):0.842089.
	 34 0.365564(4) cos:0.998754 |grad|:0.088268 |v|:0.386948 v/E(v):0.876760.
Maybe raise the learning rate a touch
	cbp> mul 2
Momentum too
	cbp> mom =
	momentum  (0 to 1)  [0.8]  .9
	momentum       	0.9
What's the learning rate?
	cbp> eps
	epsilon        	0.1
	cbp> go 10
	 35 0.365127(4) cos:0.998480 |grad|:0.087957 |v|:0.397740 v/E(v):0.452200.
	 36 0.364152(4) cos:0.996700 |grad|:0.087334 |v|:0.445816 v/E(v):0.510470.
	 37 0.363093(4) cos:0.994172 |grad|:0.086773 |v|:0.488332 v/E(v):0.562767.
	 38 0.361963(4) cos:0.990877 |grad|:0.086325 |v|:0.525849 v/E(v):0.609154.
	 39 0.360771(4) cos:0.986862 |grad|:0.086035 |v|:0.558922 v/E(v):0.649648.
	 40 0.359523(4) cos:0.982262 |grad|:0.085944 |v|:0.588099 v/E(v):0.684285.
	 41 0.358227(4) cos:0.977301 |grad|:0.086080 |v|:0.613920 v/E(v):0.713195.
	 42 0.356884(4) cos:0.972286 |grad|:0.086461 |v|:0.636915 v/E(v):0.736652.
	 43 0.355495(4) cos:0.967570 |grad|:0.087087 |v|:0.657599 v/E(v):0.755109.
	 44 0.354060(4) cos:0.963508 |grad|:0.087945 |v|:0.676459 v/E(v):0.769184.
Let's explain what these numbers mean.
	cbp> go 10
	 45 0.352577(4) cos:0.960404 |grad|:0.089010 |v|:0.693948 v/E(v):0.779627.
---------^ epoch number
	 46 0.351043(4) cos:0.958469 |grad|:0.090248 |v|:0.710472 v/E(v):0.787248.
-------------^ Error per pattern
	 47 0.349452(4) cos:0.957793 |grad|:0.091616 |v|:0.726381 v/E(v):0.792852.
---------------------^ number of patterns the net has wrong (all of them)
	 48 0.347803(4) cos:0.958344 |grad|:0.093074 |v|:0.741960 v/E(v):0.797174.
----------------------------^ cosine of angle between the
gradient and the accumulated momentum
	 49 0.346091(4) cos:0.959980 |grad|:0.094580 |v|:0.757427 v/E(v):0.800834.
--------------------------------------------^ The magnitude of the gradient
	 50 0.344311(4) cos:0.962477 |grad|:0.096098 |v|:0.772933 v/E(v):0.804316.
The magnitude of the accumulated momentum ---------------^
	 51 0.342462(4) cos:0.965571 |grad|:0.097600 |v|:0.788563 v/E(v):0.807955.
The ratio ... see display_expected_velocity -----------------------------^
	 52 0.340542(4) cos:0.968984 |grad|:0.099063 |v|:0.804348 v/E(v):0.811956.
	 53 0.338549(4) cos:0.972458 |grad|:0.100473 |v|:0.820269 v/E(v):0.816407.
	 54 0.336482(4) cos:0.975771 |grad|:0.101823 |v|:0.836275 v/E(v):0.821305.
	cbp> go 10
	 55 0.334342(4) cos:0.978755 |grad|:0.103110 |v|:0.852295 v/E(v):0.826586.
	 56 0.332130(4) cos:0.981292 |grad|:0.104338 |v|:0.868242 v/E(v):0.832144.
	 57 0.329848(4) cos:0.983322 |grad|:0.105511 |v|:0.884032 v/E(v):0.837855.
	 58 0.327495(4) cos:0.984830 |grad|:0.106637 |v|:0.899586 v/E(v):0.843598.
	 59 0.325076(4) cos:0.985842 |grad|:0.107720 |v|:0.914833 v/E(v):0.849266.
	 60 0.322591(4) cos:0.986412 |grad|:0.108768 |v|:0.929721 v/E(v):0.854774.
	 61 0.320042(4) cos:0.986616 |grad|:0.109783 |v|:0.944208 v/E(v):0.860066.
	 62 0.317431(4) cos:0.986538 |grad|:0.110768 |v|:0.958268 v/E(v):0.865115.
	 63 0.314761(4) cos:0.986264 |grad|:0.111722 |v|:0.971887 v/E(v):0.869919.
	 64 0.312033(4) cos:0.985873 |grad|:0.112643 |v|:0.985058 v/E(v):0.874499.
	cbp> go 10
	 65 0.309248(4) cos:0.985434 |grad|:0.113527 |v|:0.997782 v/E(v):0.878893.
	 66 0.306408(4) cos:0.985001 |grad|:0.114371 |v|:1.010062 v/E(v):0.883146.
	 67 0.303516(4) cos:0.984612 |grad|:0.115169 |v|:1.021901 v/E(v):0.887307.
	 68 0.300572(4) cos:0.984290 |grad|:0.115916 |v|:1.033304 v/E(v):0.891427.
	 69 0.297580(4) cos:0.984045 |grad|:0.116607 |v|:1.044269 v/E(v):0.895549.
	 70 0.294541(4) cos:0.983875 |grad|:0.117237 |v|:1.054792 v/E(v):0.899709.
	 71 0.291457(4) cos:0.983769 |grad|:0.117803 |v|:1.064866 v/E(v):0.903935.
	 72 0.288332(4) cos:0.983711 |grad|:0.118303 |v|:1.074479 v/E(v):0.908246.
	 73 0.285168(4) cos:0.983685 |grad|:0.118732 |v|:1.083615 v/E(v):0.912654.
	 74 0.281968(4) cos:0.983670 |grad|:0.119091 |v|:1.092258 v/E(v):0.917166.
	cbp> go 10
	 75 0.278736(4) cos:0.983653 |grad|:0.119376 |v|:1.100386 v/E(v):0.921784.
	 76 0.275474(4) cos:0.983622 |grad|:0.119587 |v|:1.107981 v/E(v):0.926509.
	 77 0.272188(4) cos:0.983570 |grad|:0.119721 |v|:1.115019 v/E(v):0.931345.
	 78 0.268880(4) cos:0.983497 |grad|:0.119778 |v|:1.121479 v/E(v):0.936296.
	 79 0.265554(4) cos:0.983406 |grad|:0.119755 |v|:1.127341 v/E(v):0.941374.
	 80 0.262215(4) cos:0.983308 |grad|:0.119648 |v|:1.132583 v/E(v):0.946594.
	 81 0.258867(4) cos:0.983214 |grad|:0.119455 |v|:1.137184 v/E(v):0.951978.
	 82 0.255515(4) cos:0.983140 |grad|:0.119171 |v|:1.141124 v/E(v):0.957553.
	 83 0.252162(4) cos:0.983102 |grad|:0.118792 |v|:1.144380 v/E(v):0.963352.
This is chance level, output of 0.5 for each case, error .25 for each case
	 84 0.248814(4) cos:0.983115 |grad|:0.118313 |v|:1.146933 v/E(v):0.969408.
	cbp> go 10
	 85 0.245476(4) cos:0.983192 |grad|:0.117730 |v|:1.148758 v/E(v):0.975756.
	 86 0.242152(4) cos:0.983344 |grad|:0.117040 |v|:1.149835 v/E(v):0.982430.
	 87 0.238848(4) cos:0.983576 |grad|:0.116239 |v|:1.150138 v/E(v):0.989458.
	 88 0.235569(4) cos:0.983889 |grad|:0.115327 |v|:1.149646 v/E(v):0.996862.
	 89 0.232320(4) cos:0.984280 |grad|:0.114302 |v|:1.148335 v/E(v):1.004655.
	 90 0.229106(4) cos:0.984741 |grad|:0.113165 |v|:1.146184 v/E(v):1.012840.
Starting to get it
	 91 0.225933(3) cos:0.985260 |grad|:0.111921 |v|:1.143174 v/E(v):1.021409.
	 92 0.222806(3) cos:0.985823 |grad|:0.110574 |v|:1.139289 v/E(v):1.030343.
	 93 0.219729(3) cos:0.986415 |grad|:0.109129 |v|:1.134518 v/E(v):1.039611.
	 94 0.216707(3) cos:0.987019 |grad|:0.107595 |v|:1.128855 v/E(v):1.049171.
	cbp> go 50
	 95 0.213745(3) cos:0.987619 |grad|:0.105980 |v|:1.122301 v/E(v):1.058974.
	 96 0.210847(3) cos:0.988201 |grad|:0.104294 |v|:1.114862 v/E(v):1.068961.
	 97 0.208016(3) cos:0.988750 |grad|:0.102547 |v|:1.106555 v/E(v):1.079066.
	 98 0.205255(3) cos:0.989255 |grad|:0.100751 |v|:1.097400 v/E(v):1.089222.
falling into place
	 99 0.202566(2) cos:0.989708 |grad|:0.098915 |v|:1.087428 v/E(v):1.099354.
	 100 0.199953(2) cos:0.990100 |grad|:0.097051 |v|:1.076676 v/E(v):1.109389.
	 101 0.197416(2) cos:0.990427 |grad|:0.095170 |v|:1.065186 v/E(v):1.119251.
	 102 0.194957(2) cos:0.990685 |grad|:0.093280 |v|:1.053008 v/E(v):1.128863.
	 103 0.192576(2) cos:0.990871 |grad|:0.091393 |v|:1.040196 v/E(v):1.138153.
	 104 0.190273(2) cos:0.990984 |grad|:0.089518 |v|:1.026809 v/E(v):1.147047.
	 105 0.188048(2) cos:0.991023 |grad|:0.087662 |v|:1.012910 v/E(v):1.155474.
	 106 0.185901(2) cos:0.990987 |grad|:0.085834 |v|:0.998563 v/E(v):1.163368.
	 107 0.183829(2) cos:0.990877 |grad|:0.084041 |v|:0.983834 v/E(v):1.170663.
	 108 0.181832(2) cos:0.990693 |grad|:0.082289 |v|:0.968791 v/E(v):1.177299.
	 109 0.179907(2) cos:0.990435 |grad|:0.080585 |v|:0.953501 v/E(v):1.183219.
	 110 0.178053(2) cos:0.990104 |grad|:0.078934 |v|:0.938031 v/E(v):1.188371.
	 111 0.176268(2) cos:0.989700 |grad|:0.077341 |v|:0.922448 v/E(v):1.192707.
	 112 0.174549(2) cos:0.989222 |grad|:0.075809 |v|:0.906815 v/E(v):1.196184.
	 113 0.172893(2) cos:0.988673 |grad|:0.074343 |v|:0.891194 v/E(v):1.198764.
	 114 0.171298(2) cos:0.988051 |grad|:0.072945 |v|:0.875646 v/E(v):1.200415.
	 115 0.169762(2) cos:0.987359 |grad|:0.071620 |v|:0.860229 v/E(v):1.201111.
	 116 0.168280(2) cos:0.986599 |grad|:0.070368 |v|:0.844996 v/E(v):1.200831.
	 117 0.166850(2) cos:0.985772 |grad|:0.069192 |v|:0.830001 v/E(v):1.199562.
	 118 0.165470(2) cos:0.984882 |grad|:0.068094 |v|:0.815291 v/E(v):1.197296.
	 119 0.164136(2) cos:0.983934 |grad|:0.067076 |v|:0.800914 v/E(v):1.194035.
	 120 0.162846(2) cos:0.982934 |grad|:0.066139 |v|:0.786912 v/E(v):1.189787.
	 121 0.161596(2) cos:0.981890 |grad|:0.065283 |v|:0.773327 v/E(v):1.184569.
	 122 0.160384(2) cos:0.980810 |grad|:0.064511 |v|:0.760196 v/E(v):1.178405.
	 123 0.159207(2) cos:0.979706 |grad|:0.063821 |v|:0.747555 v/E(v):1.171330.
	 124 0.158061(2) cos:0.978590 |grad|:0.063215 |v|:0.735437 v/E(v):1.163386.
	 125 0.156945(2) cos:0.977477 |grad|:0.062693 |v|:0.723872 v/E(v):1.154624.
	 126 0.155855(2) cos:0.976381 |grad|:0.062255 |v|:0.712888 v/E(v):1.145105.
	 127 0.154789(2) cos:0.975319 |grad|:0.061901 |v|:0.702513 v/E(v):1.134898.
	 128 0.153744(2) cos:0.974308 |grad|:0.061630 |v|:0.692770 v/E(v):1.124077.
	 129 0.152718(2) cos:0.973366 |grad|:0.061442 |v|:0.683681 v/E(v):1.112726.
	 130 0.151707(2) cos:0.972509 |grad|:0.061336 |v|:0.675265 v/E(v):1.100934.
	 131 0.150710(2) cos:0.971755 |grad|:0.061310 |v|:0.667541 v/E(v):1.088793.
	 132 0.149723(2) cos:0.971119 |grad|:0.061364 |v|:0.660524 v/E(v):1.076401.
	 133 0.148745(2) cos:0.970613 |grad|:0.061496 |v|:0.654227 v/E(v):1.063856.
	 134 0.147773(2) cos:0.970248 |grad|:0.061703 |v|:0.648662 v/E(v):1.051257.
	 135 0.146805(2) cos:0.970034 |grad|:0.061985 |v|:0.643836 v/E(v):1.038700.
	 136 0.145838(2) cos:0.969974 |grad|:0.062338 |v|:0.639757 v/E(v):1.026281.
	 137 0.144870(2) cos:0.970070 |grad|:0.062759 |v|:0.636428 v/E(v):1.014088.
	 138 0.143898(2) cos:0.970322 |grad|:0.063245 |v|:0.633849 v/E(v):1.002208.
	 139 0.142921(2) cos:0.970723 |grad|:0.063794 |v|:0.632017 v/E(v):0.990717.
	 140 0.141937(2) cos:0.971267 |grad|:0.064401 |v|:0.630928 v/E(v):0.979687.
	 141 0.140943(2) cos:0.971941 |grad|:0.065063 |v|:0.630572 v/E(v):0.969178.
	 142 0.139936(2) cos:0.972734 |grad|:0.065774 |v|:0.630937 v/E(v):0.959246.
	 143 0.138916(2) cos:0.973630 |grad|:0.066532 |v|:0.632009 v/E(v):0.949935.
	 144 0.137881(2) cos:0.974614 |grad|:0.067330 |v|:0.633767 v/E(v):0.941281.
	cbp> go 100
	 145 0.136827(2) cos:0.975669 |grad|:0.068165 |v|:0.636190 v/E(v):0.933314.
	 146 0.135754(2) cos:0.976777 |grad|:0.069030 |v|:0.639252 v/E(v):0.926053.
	 147 0.134660(2) cos:0.977923 |grad|:0.069920 |v|:0.642923 v/E(v):0.919513.
	 148 0.133544(2) cos:0.979091 |grad|:0.070830 |v|:0.647172 v/E(v):0.913701.
	 149 0.132403(2) cos:0.980266 |grad|:0.071753 |v|:0.651963 v/E(v):0.908617.
	 150 0.131236(2) cos:0.981435 |grad|:0.072685 |v|:0.657257 v/E(v):0.904260.
	 151 0.130043(2) cos:0.982586 |grad|:0.073618 |v|:0.663013 v/E(v):0.900620.
	 152 0.128822(2) cos:0.983709 |grad|:0.074546 |v|:0.669187 v/E(v):0.897686.
	 153 0.127573(2) cos:0.984795 |grad|:0.075464 |v|:0.675733 v/E(v):0.895443.
	 154 0.126295(2) cos:0.985838 |grad|:0.076364 |v|:0.682601 v/E(v):0.893876.
	 155 0.124986(2) cos:0.986832 |grad|:0.077242 |v|:0.689743 v/E(v):0.892965.
	 156 0.123648(2) cos:0.987773 |grad|:0.078090 |v|:0.697105 v/E(v):0.892691.
	 157 0.122280(2) cos:0.988658 |grad|:0.078904 |v|:0.704636 v/E(v):0.893032.
	 158 0.120881(2) cos:0.989486 |grad|:0.079676 |v|:0.712279 v/E(v):0.893966.
	 159 0.119454(2) cos:0.990257 |grad|:0.080403 |v|:0.719982 v/E(v):0.895472.
	 160 0.117997(2) cos:0.990970 |grad|:0.081077 |v|:0.727689 v/E(v):0.897526.
	 161 0.116512(2) cos:0.991627 |grad|:0.081696 |v|:0.735346 v/E(v):0.900106.
	 162 0.115001(2) cos:0.992230 |grad|:0.082253 |v|:0.742898 v/E(v):0.903188.
	 163 0.113463(2) cos:0.992780 |grad|:0.082745 |v|:0.750291 v/E(v):0.906750.
	 164 0.111902(2) cos:0.993281 |grad|:0.083169 |v|:0.757475 v/E(v):0.910768.
	 165 0.110319(2) cos:0.993734 |grad|:0.083521 |v|:0.764398 v/E(v):0.915220.
	 166 0.108715(2) cos:0.994144 |grad|:0.083798 |v|:0.771012 v/E(v):0.920080.
	 167 0.107092(1) cos:0.994512 |grad|:0.084000 |v|:0.777271 v/E(v):0.925327.
	 168 0.105454(1) cos:0.994842 |grad|:0.084123 |v|:0.783132 v/E(v):0.930936.
	 169 0.103803(1) cos:0.995137 |grad|:0.084168 |v|:0.788554 v/E(v):0.936883.
	 170 0.102140(1) cos:0.995400 |grad|:0.084134 |v|:0.793501 v/E(v):0.943144.
	 171 0.100470(1) cos:0.995633 |grad|:0.084021 |v|:0.797938 v/E(v):0.949693.
	 172 0.098794(1) cos:0.995838 |grad|:0.083830 |v|:0.801836 v/E(v):0.956504.
	 173 0.097115(1) cos:0.996018 |grad|:0.083563 |v|:0.805170 v/E(v):0.963553.
	 174 0.095437(1) cos:0.996175 |grad|:0.083221 |v|:0.807917 v/E(v):0.970810.
	 175 0.093761(1) cos:0.996310 |grad|:0.082807 |v|:0.810060 v/E(v):0.978250.
	 176 0.092091(1) cos:0.996424 |grad|:0.082324 |v|:0.811587 v/E(v):0.985843.
	 177 0.090430(1) cos:0.996519 |grad|:0.081775 |v|:0.812488 v/E(v):0.993560.
	 178 0.088779(1) cos:0.996596 |grad|:0.081164 |v|:0.812758 v/E(v):1.001373.
	 179 0.087142(1) cos:0.996655 |grad|:0.080495 |v|:0.812398 v/E(v):1.009250.
	 180 0.085522(1) cos:0.996697 |grad|:0.079772 |v|:0.811411 v/E(v):1.017161.
took long enough (parameters used were very conservative, and initial
weights were too big)
	 181 0.083919(0) cos:0.996723 |grad|:0.079000 |v|:0.809805 v/E(v):1.025076.
	 182 0.082337(0) cos:0.996734 |grad|:0.078182 |v|:0.807590 v/E(v):1.032962.
	 183 0.080777(0) cos:0.996729 |grad|:0.077324 |v|:0.804782 v/E(v):1.040791.
	 184 0.079242(0) cos:0.996711 |grad|:0.076431 |v|:0.801400 v/E(v):1.048532.
	 185 0.077732(0) cos:0.996679 |grad|:0.075506 |v|:0.797463 v/E(v):1.056155.
	 186 0.076249(0) cos:0.996636 |grad|:0.074556 |v|:0.792996 v/E(v):1.063632.
	 187 0.074794(0) cos:0.996581 |grad|:0.073583 |v|:0.788025 v/E(v):1.070937.
	 188 0.073368(0) cos:0.996517 |grad|:0.072592 |v|:0.782577 v/E(v):1.078043.
	 189 0.071973(0) cos:0.996445 |grad|:0.071588 |v|:0.776682 v/E(v):1.084927.
	 190 0.070608(0) cos:0.996365 |grad|:0.070575 |v|:0.770372 v/E(v):1.091567.
	 191 0.069275(0) cos:0.996282 |grad|:0.069555 |v|:0.763676 v/E(v):1.097944.
	 192 0.067974(0) cos:0.996195 |grad|:0.068533 |v|:0.756629 v/E(v):1.104041.
	 193 0.066704(0) cos:0.996107 |grad|:0.067511 |v|:0.749262 v/E(v):1.109842.
	 194 0.065466(0) cos:0.996019 |grad|:0.066492 |v|:0.741607 v/E(v):1.115335.
	 195 0.064260(0) cos:0.995933 |grad|:0.065479 |v|:0.733698 v/E(v):1.120512.
	 196 0.063086(0) cos:0.995851 |grad|:0.064474 |v|:0.725565 v/E(v):1.125364.
	 197 0.061943(0) cos:0.995775 |grad|:0.063479 |v|:0.717238 v/E(v):1.129886.
	 198 0.060832(0) cos:0.995705 |grad|:0.062496 |v|:0.708749 v/E(v):1.134077.
	 199 0.059750(0) cos:0.995642 |grad|:0.061526 |v|:0.700125 v/E(v):1.137936.
	 200 0.058699(0) cos:0.995589 |grad|:0.060571 |v|:0.691394 v/E(v):1.141465.
	 201 0.057678(0) cos:0.995545 |grad|:0.059631 |v|:0.682582 v/E(v):1.144668.
	 202 0.056686(0) cos:0.995511 |grad|:0.058709 |v|:0.673713 v/E(v):1.147551.
	 203 0.055721(0) cos:0.995488 |grad|:0.057804 |v|:0.664810 v/E(v):1.150121.
	 204 0.054785(0) cos:0.995476 |grad|:0.056916 |v|:0.655895 v/E(v):1.152386.
	 205 0.053875(0) cos:0.995474 |grad|:0.056047 |v|:0.646987 v/E(v):1.154357.
	 206 0.052991(0) cos:0.995484 |grad|:0.055197 |v|:0.638104 v/E(v):1.156045.
	 207 0.052133(0) cos:0.995505 |grad|:0.054366 |v|:0.629263 v/E(v):1.157461.
	 208 0.051299(0) cos:0.995536 |grad|:0.053553 |v|:0.620480 v/E(v):1.158619.
	 209 0.050490(0) cos:0.995576 |grad|:0.052760 |v|:0.611767 v/E(v):1.159530.
	 210 0.049703(0) cos:0.995626 |grad|:0.051985 |v|:0.603137 v/E(v):1.160209.
	 211 0.048939(0) cos:0.995685 |grad|:0.051229 |v|:0.594601 v/E(v):1.160669.
Let it keep running for a while to firm things up
	 212 0.048197(0) cos:0.995752 |grad|:0.050492 |v|:0.586168 v/E(v):1.160923.
	 213 0.047475(0) cos:0.995826 |grad|:0.049772 |v|:0.577847 v/E(v):1.160986.
	 214 0.046774(0) cos:0.995907 |grad|:0.049071 |v|:0.569645 v/E(v):1.160871.
	 215 0.046093(0) cos:0.995993 |grad|:0.048386 |v|:0.561568 v/E(v):1.160590.
	 216 0.045430(0) cos:0.996085 |grad|:0.047719 |v|:0.553620 v/E(v):1.160158.
	 217 0.044785(0) cos:0.996181 |grad|:0.047069 |v|:0.545807 v/E(v):1.159586.
	 218 0.044159(0) cos:0.996280 |grad|:0.046435 |v|:0.538132 v/E(v):1.158886.
	 219 0.043549(0) cos:0.996383 |grad|:0.045817 |v|:0.530596 v/E(v):1.158071.
	 220 0.042955(0) cos:0.996487 |grad|:0.045215 |v|:0.523203 v/E(v):1.157151.
	 221 0.042378(0) cos:0.996593 |grad|:0.044627 |v|:0.515952 v/E(v):1.156136.
	 222 0.041816(0) cos:0.996700 |grad|:0.044054 |v|:0.508845 v/E(v):1.155038.
	 223 0.041268(0) cos:0.996807 |grad|:0.043496 |v|:0.501883 v/E(v):1.153864.
	 224 0.040735(0) cos:0.996915 |grad|:0.042951 |v|:0.495063 v/E(v):1.152624.
	 225 0.040215(0) cos:0.997021 |grad|:0.042420 |v|:0.488387 v/E(v):1.151327.
	 226 0.039708(0) cos:0.997127 |grad|:0.041901 |v|:0.481853 v/E(v):1.149979.
	 227 0.039215(0) cos:0.997231 |grad|:0.041395 |v|:0.475459 v/E(v):1.148588.
	 228 0.038733(0) cos:0.997333 |grad|:0.040901 |v|:0.469203 v/E(v):1.147160.
	 229 0.038263(0) cos:0.997433 |grad|:0.040419 |v|:0.463085 v/E(v):1.145702.
	 230 0.037805(0) cos:0.997532 |grad|:0.039949 |v|:0.457101 v/E(v):1.144219.
	 231 0.037358(0) cos:0.997627 |grad|:0.039489 |v|:0.451250 v/E(v):1.142717.
	 232 0.036922(0) cos:0.997721 |grad|:0.039040 |v|:0.445529 v/E(v):1.141200.
	 233 0.036496(0) cos:0.997811 |grad|:0.038602 |v|:0.439935 v/E(v):1.139672.
	 234 0.036080(0) cos:0.997899 |grad|:0.038174 |v|:0.434466 v/E(v):1.138138.
	 235 0.035673(0) cos:0.997983 |grad|:0.037755 |v|:0.429120 v/E(v):1.136600.
	 236 0.035276(0) cos:0.998065 |grad|:0.037345 |v|:0.423894 v/E(v):1.135063.
	 237 0.034888(0) cos:0.998144 |grad|:0.036945 |v|:0.418784 v/E(v):1.133529.
	 238 0.034508(0) cos:0.998220 |grad|:0.036554 |v|:0.413788 v/E(v):1.132000.
	 239 0.034137(0) cos:0.998293 |grad|:0.036171 |v|:0.408904 v/E(v):1.130479.
	 240 0.033774(0) cos:0.998363 |grad|:0.035796 |v|:0.404128 v/E(v):1.128968.
	 241 0.033419(0) cos:0.998430 |grad|:0.035430 |v|:0.399458 v/E(v):1.127469.
	 242 0.033071(0) cos:0.998495 |grad|:0.035071 |v|:0.394891 v/E(v):1.125984.
	 243 0.032731(0) cos:0.998557 |grad|:0.034719 |v|:0.390425 v/E(v):1.124514.
	 244 0.032398(0) cos:0.998616 |grad|:0.034375 |v|:0.386056 v/E(v):1.123060.
Okay, now things should be pretty firm.  Let's look at the outputs
	cbp> write_st
	State file to write: /dev/tty
	11 units
	4 patterns
	1 1 0 0 0 0.057084 0.036367 0.791040 0.168167 0.006977 0.147379 
	1 0 1 0 0 0.923575 0.041349 0.139184 0.820154 0.148581 0.002259 
	1 0 0 1 0 0.961232 0.959144 0.003733 0.107678 0.835538 0.063689 
	1 0 0 0 1 0.026562 0.921445 0.113286 0.004758 0.124165 0.865541 
inputs  \-------/
hidden units      \---------------/
output units                         \--------------------------------/
note that the first input is really just the "bias unit"
	cbp> 
	cbp> ^D









			External File Formats

Weight files (.wet) have the following format:

	<n> epochs
	<n> weights
	<w1> <w2> ... <wn>
      [ <n> timeconstants
	<t1> <t2> ... <tn> ]

Network descriptor files (.net) have the following format:

	<n> units
	<n> inputs
	<u>^n
	<n> outputs
	<u>^n
	<n> conns
	(<u1> <u2>)^n
      [ : <n> equivalences
	(<m> <l>^m)^n ]
      [ t
	<transfer-function>^n ]


Environment descriptor files (.env) have the following format:

	<n> inputs
	<n> outputs
	<n> patterns
	<patt>^n
      [ : <m> extrasteps
        ( t=<t>
	  <patt>^n )^m ]

    where <patt> can take one of two forms

	<r>^n
	: ( <r> | ?<r> )^inputs ( <r> | * )^outputs
	: ( <r> | ?<r> )^(inputs+outputs)

State files have the following format:

	<n> units
	<m> patterns
	(<r>^n)^m