-
Notifications
You must be signed in to change notification settings - Fork 1
How to run the program
Juan Lao Tebar edited this page Jan 19, 2017
·
3 revisions
This page describes how to execute the generator of results of this project.
In order to execute the generator, open a system shell and go to the src directory of this project. To see a full list of available options, execute the generator with the -h option:
$ cd src
$ python -m bblr.Main -h
usage: bblr-hopfield-boltzmann analyzer [-h] [--seed SEED]
[--just-analyze-patterns]
[--just-analyze-inputs]
[--out RESULTS_FILE]
pattern_data_set_properties_file
model_properties_file
input_data_set_properties_file
Automatic script to obtain the final results of the project bblr-hopfield-
boltzmann.
positional arguments:
pattern_data_set_properties_file
Path to a file containing the combinations of pattern
data set properties you want to test. See
https://github.com/juanlao7/bblr-hopfield-
boltzmann/wiki/Pattern-data-set-properties-file-format
model_properties_file
Path to a file containing the combinations of model
properties you want to test. See
https://github.com/juanlao7/bblr-hopfield-
boltzmann/wiki/Pattern-data-set-properties-file-format
input_data_set_properties_file
Path to a file containing the combinations of input
data set properties you want to test. See
https://github.com/juanlao7/bblr-hopfield-
boltzmann/wiki/Input-data-set-properties-file-format
optional arguments:
-h, --help show this help message and exit
--seed SEED This random seed is added to the seed defined in the
configuration files to obtain different results.
--just-analyze-patterns
Just create the random pattern data sets and analyze
them. Do not do anything else.
--just-analyze-inputs
Just create the random pattern data sets, the random
input data sets and analyze them. Do not do anything
else.
--out RESULTS_FILE Store the results as a JSON file.Before getting results, you would like to check if your pattern data sets are being generated correctly. This is interesting, since this generator approximates the configured distance between patterns, and the generated pattern data set may not have the desired values.
To analyze the pattern data sets, use the --just-analyze-patterns option:
$ cd src
$ python -m bblr.Main --just-analyze-patterns ../config/pattern/final.json ../config/model/final.json ../config/input/final.json
PATTERN: {u'dataSetSize': 30, u'distance': {u'stdev': 2, u'mean': 15}, u'patternSize': 25, u'seed': 0}
Pattern data set size : 30
Pattern dimension : 25
Distance between patterns mean : 12.9149425287
Distance between patterns standard deviation : 2.30932784845
PATTERN: {u'dataSetSize': 30, u'distance': {u'stdev': 2, u'mean': 15}, u'patternSize': 25, u'seed': 0, u'extraBits': {u'values': u'random', u'number': 25}}
Pattern data set size : 30
Pattern dimension : 50
Distance between patterns mean : 25.4643678161
Distance between patterns standard deviation : 3.42411921761Before getting results, you would like to check if your input data sets are being generated correctly.
To analyze the input data sets, use the --just-analyze-inputs option:
$ cd src
$ python -m bblr.Main --just-analyze-inputs ../config/pattern/final.json ../config/model/final.json ../config/input/final.json
PATTERN: {u'dataSetSize': 30, u'distance': {u'stdev': 2, u'mean': 15}, u'patternSize': 25, u'seed': 0}
Pattern data set size : 30
Pattern dimension : 25
Distance between patterns mean : 12.9149425287
Distance between patterns standard deviation : 2.30932784845
INPUT: {u'inputsPerPattern': 3, u'seed': 0, u'minDistance': {u'stdev': 2, u'mean': 3}}
Input data set size : 90.0
Input vector dimension : 25
Minimum distance of inputs to patterns mean : 3.9
Minimum distance of inputs to patterns standard deviation : 1.88591268798
INPUT: {u'inputsPerPattern': 3, u'seed': 0, u'minDistance': {u'stdev': 2, u'mean': 3}}
Input data set size : 90.0
Input vector dimension : 25
Minimum distance of inputs to patterns mean : 3.9
Minimum distance of inputs to patterns standard deviation : 1.88591268798
PATTERN: {u'dataSetSize': 30, u'distance': {u'stdev': 2, u'mean': 15}, u'patternSize': 25, u'seed': 0, u'extraBits': {u'values': u'random', u'number': 25}}
Pattern data set size : 30
Pattern dimension : 50
Distance between patterns mean : 25.4643678161
Distance between patterns standard deviation : 3.42411921761
INPUT: {u'inputsPerPattern': 3, u'seed': 0, u'minDistance': {u'stdev': 2, u'mean': 3}}
Input data set size : 90.0
Input vector dimension : 25
Minimum distance of inputs to patterns mean : 3.9
Minimum distance of inputs to patterns standard deviation : 1.88591268798
INPUT: {u'inputsPerPattern': 3, u'seed': 0, u'minDistance': {u'stdev': 2, u'mean': 3}}
Input data set size : 90.0
Input vector dimension : 25
Minimum distance of inputs to patterns mean : 3.9
Minimum distance of inputs to patterns standard deviation : 1.88591268798As you can see, pattern data sets are also analyzed, since input data sets depend strongly on pattern data sets and this information is relevant too.
To generate an output JSON file with the computed results after testing all the configured combinations, execute the generator like this:
$ cd src
$ python -m bblr.Main --out ../output.json ../config/pattern/final.json ../config/model/final.json ../config/input/final.json
PATTERN: {u'dataSetSize': 30, u'distance': {u'stdev': 2, u'mean': 15}, u'patternSize': 25, u'seed': 0}
Pattern data set size : 30
Pattern dimension : 25
Distance between patterns mean : 12.9149425287
Distance between patterns standard deviation : 2.30932784845
MODEL: {u'model': u'hopfield', u'trainingRule': u'hebbian'}
INPUT: {u'inputsPerPattern': 3, u'seed': 0, u'minDistance': {u'stdev': 2, u'mean': 3}}
Input data set size : 90.0
Input vector dimension : 25
Minimum distance of inputs to patterns mean : 3.9
Minimum distance of inputs to patterns standard deviation : 1.88591268798
RESULT:
CPU time mean : 0.277777777778
CPU time standard deviation : 0.447903208239
Iterations mean : 2.0
Iterations standard deviation : 0.0
Successful equilibriums : 0
Unsuccessful equilibriums : 90
Spurious equilibriums : 0
[...]