def visualize_correctness(n=25, working_dir=None):
if not working_dir:
working_dir = get_working_dir()
(reservoir_input, expected_output), _ =\
glob_load(working_dir + '*-dataset')[0]
rbn_reservoir, _ = glob_load(working_dir + '*-reservoir')[0]
readout, _ = glob_load(working_dir + '*-readout')[0]
rbn_reservoir.reset_state()
flow = mdp.Flow([rbn_reservoir, readout], verbose=1)
actual_output = flow.execute(reservoir_input)
for output in actual_output:
output[0] = 1 if output[0] > 0.5 else 0
errors = sum(actual_output != expected_output)
accuracy = 1 - float(errors) / len(actual_output)
plt.title('Reservoir performance')
plt.plot(actual_output[:n], 'y', linewidth=1.5)
plt.plot(expected_output[:n], 'b', linewidth=1.5)
plt.legend(['Actual output', 'Expected output'])
plt.savefig('temp-2.pdf', bbox_inches='tight')
def load_rbns_from_erb(directory, remember):
rbn_distribution, name = glob_load(directory + '*distribution')[0]
scatterplot = ["x y"]
for accuracy, rbn in rbn_distribution:
cc = measure_computational_capability(rbn, 100, remember)
scatterplot.append("{} {}".format(cc, accuracy))
distribution = {}
distplot = ["\\myboxplot{"]
for accuracy, rbn in rbn_distribution:
ic = rbn.input_connectivity
if ic not in distribution:
distribution[ic] = []
distribution[ic].append(accuracy)
for i, l in enumerate(sorted(distribution.keys())):
distplot.append("% L: {}".format(i))
distplot.append("\\addplot[boxplot]")
distplot.append("table[row sep=\\\\, y index=0] {")
distplot.append("data")
for fitness in distribution[l]:
distplot.append("{} \\\\".format(fitness))
distplot.append("};")
distplot.append("}}{{{}}}".format(10.0 / max(distribution.keys())))
return '\n'.join(scatterplot), '\n'.join(distplot)
def load_rbns_from_erb(directory, remember):
rbn_distribution, name = glob_load(directory + '*distribution')[0]
scatterplot = ["x y"]
for accuracy, rbn in rbn_distribution:
cc = measure_computational_capability(rbn, 100, remember)
scatterplot.append("{} {}".format(cc, accuracy))
distribution = {}
distplot = ["\\myboxplot{"]
for accuracy, rbn in rbn_distribution:
ic = rbn.input_connectivity
if ic not in distribution:
distribution[ic] = []
distribution[ic].append(accuracy)
for i, l in enumerate(sorted(distribution.keys())):
distplot.append("% L: {}".format(i))
distplot.append("\\addplot[boxplot]")
distplot.append("table[row sep=\\\\, y index=0] {")
distplot.append("data")
for fitness in distribution[l]:
distplot.append("{} \\\\".format(fitness))
distplot.append("};")
distplot.append("}}{{{}}}".format(10.0 / max(distribution.keys())))
return '\n'.join(scatterplot), '\n'.join(distplot)
def load_rbns_from_ea():
working_dir = get_working_dir()
ea_runs = map(fst, glob_load(working_dir + '*-evolved'))
best_genomes = map(lst, ea_runs)
rbns = [genotype_to_phenotype(genome, 100, 2) for genome in best_genomes]
return best_genomes, rbns
def load_rbns_from_ea():
working_dir = get_working_dir()
ea_runs = map(fst, glob_load(working_dir + '*-evolved'))
best_genomes = map(lst, ea_runs)
rbns = [genotype_to_phenotype(genome, 100, 2)
for genome in best_genomes]
return best_genomes, rbns
def visualize_rbn_state(n=100, working_dir=None):
if not working_dir:
working_dir = get_working_dir()
rbn, rbn_name = glob_load(working_dir + '*-reservoir')[0]
rbn.reset_state()
if not user_denies('Perturb?'):
test_data, _ = glob_load(working_dir + '*-dataset')[0]
test_input, _ = test_data
test_input = test_input[:n]
else:
test_input = np.zeros((n, 1))
rbn.should_perturb = False
rbn_states = rbn._execute(test_input)
plt.matshow(rbn_states, cmap=plt.cm.gray)
plt.axis('off')
#plt.xlabel('State of node n in RBN')
#plt.gca().xaxis.set_label_position('top')
#plt.ylabel('Time')
plt.savefig(raw_input('Name: '), bbox_inches='tight')
def visualize_dataset(n=30, working_dir=None):
if not working_dir:
working_dir = get_working_dir()
test_dataset, filename = glob_load(working_dir + '*-dataset')[0]
dataset_meta = re.search(r"\[(.*)\]", filename).groups()[0]
reservoir_input = np.transpose(test_dataset[0][:n])
expected_output = np.transpose(test_dataset[1][:n])
plt.matshow(reservoir_input, cmap=plt.cm.gray)
plt.axis('off')
plt.savefig('plots/' + dataset_meta + '-input.pdf', bbox_inches='tight')
plt.matshow(expected_output, cmap=plt.cm.gray)
plt.axis('off')
plt.savefig('plots/' + dataset_meta + '-output.pdf', bbox_inches='tight')
plt.show()
# lower whisker={}
#}},
#] coordinates {{}};""".format(l, i, median, upperq, lowerq, upperw, lowerw)
#
# print boxplot
#
# print "}}{{{}}}".format(10.0 / max(distribution.keys()))
if __name__ == '__main__':
#filename = "pickle_dumps/distribution-100-5-3/combined-distribution"
#computational_power_scatter()
#distribution_to_plot()
#erb()
rbn, _ = glob_load('pickle_dumps/70input-2/*-reservoir')[0]
ccs = [measure_computational_capability(rbn, 100, 3) for _ in range(20)]
print ccs, np.median(ccs), np.mean(ccs)
#postfix = default_input('Postfix?', '-3-1')
#remember = int(postfix.split("-")[1]) - 1
#scatter, box = load_rbns_from_erb('pickle_dumps/c-distribution-100' + postfix + '/', remember)
#print scatter
#sc = open('pickle_dumps/c-distribution-100' + postfix +
# '/computational-power-100' + postfix + '.dat', 'w')
#sc.write(scatter)
#sc.close()
# upper whisker={},
# lower whisker={}
#}},
#] coordinates {{}};""".format(l, i, median, upperq, lowerq, upperw, lowerw)
#
# print boxplot
#
# print "}}{{{}}}".format(10.0 / max(distribution.keys()))
if __name__ == '__main__':
#filename = "pickle_dumps/distribution-100-5-3/combined-distribution"
#computational_power_scatter()
#distribution_to_plot()
#erb()
rbn, _ = glob_load('pickle_dumps/70input-2/*-reservoir')[0]
ccs = [measure_computational_capability(rbn, 100, 3) for _ in range(20)]
print ccs, np.median(ccs), np.mean(ccs)
#postfix = default_input('Postfix?', '-3-1')
#remember = int(postfix.split("-")[1]) - 1
#scatter, box = load_rbns_from_erb('pickle_dumps/c-distribution-100' + postfix + '/', remember)
#print scatter
#sc = open('pickle_dumps/c-distribution-100' + postfix +
# '/computational-power-100' + postfix + '.dat', 'w')
#sc.write(scatter)
#sc.close()
input_connectivity=input_connectivity)
return rbn_reservoir
if __name__ == '__main__':
# Set pickle working dir
working_dir = get_working_dir()
log.setup(logging.DEBUG, path=working_dir)
log_git_info()
# Create datasets
use_existing_dataset = user_confirms('Use existing dataset in folder?')
if use_existing_dataset:
test_dataset, _ = glob_load(working_dir + '*-dataset')[0]
dataset_description = '[dataset_from_folder]'
else:
datasets, dataset_description = create_dataset()
training_dataset, test_dataset = datasets[:-1], datasets[-1]
if not use_existing_dataset and not user_denies('Pickle test dataset?'):
dump(test_dataset, dataset_description + '-dataset',
folder=working_dir)
# Create or load reservoir and readout layer
if user_confirms('Use readout layer from folder?'):
readout, _ = glob_load(working_dir + '*readout')[0]
else:
rbn_reservoir = create_reservoir()
readout = Oger.nodes.RidgeRegressionNode(
input_connectivity=input_connectivity)
return rbn_reservoir
if __name__ == '__main__':
# Set pickle working dir
working_dir = get_working_dir()
log.setup(logging.DEBUG, path=working_dir)
log_git_info()
# Create datasets
use_existing_dataset = user_confirms('Use existing dataset in folder?')
if use_existing_dataset:
test_dataset, _ = glob_load(working_dir + '*-dataset')[0]
dataset_description = '[dataset_from_folder]'
else:
datasets, dataset_description = create_dataset()
training_dataset, test_dataset = datasets[:-1], datasets[-1]
if not use_existing_dataset and not user_denies('Pickle test dataset?'):
dump(test_dataset,
dataset_description + '-dataset',
folder=working_dir)
# Create or load reservoir and readout layer
if user_confirms('Use readout layer from folder?'):
readout, _ = glob_load(working_dir + '*readout')[0]
else:
rbn_reservoir = create_reservoir()