def run_classify(context, labels_test, sample_duration_test):
# Uses outputs as inputs for the matrix! This is because the weights are symmetrized and written in the output.
os.system('rm -rf outputs/{directory}/test'.format(**context))
os.system('mkdir -p outputs/{directory}/test/'.format(**context))
ret = os.system('mpirun -n {ncores} ./exp_rbp_flash \
--learn false \
--eta 0.\
--simtime {simtime_test} \
--record_full false \
--record_rasters false \
--record_rates true \
--dir outputs/{directory}/test/ \
--fvh inputs/{directory}/train/{fvh} \
--fho inputs/{directory}/train/{fho} \
--fhh inputs/{directory}/train/{fhh} \
--foe inputs/{directory}/train/{foe} \
--feo inputs/{directory}/train/{feo} \
--fve inputs/{directory}/train/{fve} \
--feh inputs/{directory}/train/{feh} \
--ip_v inputs/{directory}/test/{ip_v}\
--prob_syn 1.\
--nvis {nv} \
--nhid {nh} \
--nout {nc} \
--sigma {sigma}\
'.format(**context))
if ret == 0:
print('ran')
plotter.plot_ras_spikes('outputs/{}/test/coba.*.{}.ras'.format(context['directory'], '{}'), start=0, end=15,
layers=['out'], res=32 * 32, number_of_classes=context['nc'], save=True)
# first 5 labels: 7,2,1,0,4
return elib.process_test_classification(context, sample_duration_test, labels_test)
def run_classify(context, labels_test, sample_duration_test):
# Uses outputs as inputs for the matrix! This is because the weights are symmetrized and written in the output.
os.system('rm -rf outputs/{directory}/test'.format(**context))
os.system('mkdir -p outputs/{directory}/test/'.format(**context))
ret = os.system('mpirun -n {ncores} ./exp_rbp_flash \
--learn false \
--eta 0.\
--simtime {simtime_test} \
--record_full false \
--record_rasters false \
--record_rates true \
--dir outputs/{directory}/test/ \
--fvh inputs/{directory}/train/{fvh} \
--fho inputs/{directory}/train/{fho} \
--fhh inputs/{directory}/train/{fhh} \
--foe inputs/{directory}/train/{foe} \
--feo inputs/{directory}/train/{feo} \
--fve inputs/{directory}/train/{fve} \
--feh inputs/{directory}/train/{feh} \
--ip_v inputs/{directory}/test/{ip_v}\
--prob_syn {prob_syn}\
--nvis {nv} \
--nhid {nh} \
--nout {nc} \
--sigma {sigma}\
'.format(**context))
if ret == 0:
print('ran')
plotter.plot_ras_spikes('outputs/{}/test/coba.*.{}.ras'.format(
context['directory'], '{}'),
start=0,
end=15,
layers=['out'],
res=context['nv'] - context['nc'],
number_of_classes=context['nc'],
save=True)
# first 5 labels: 7,2,1,0,4
rate_class, first_class, rate_confusion_data_frame, first_confusion_data_frame, ouput_spikes_per_label, ouput_spikes_per_label_norm = elib.process_test_classification(
context, sample_duration_test, labels_test)
plotter.plot_output_spike_count(ouput_spikes_per_label,
'Output spike count per label',
save=True,
image_title='out_spk')
plotter.plot_output_spike_count(ouput_spikes_per_label_norm,
'Normalized ouput spike count per label',
save=True,
image_title='out_spk_norm')
plotter.plot_confusion_matrix(rate_confusion_data_frame, save=True)
#plotter.plot_confusion_matrix(first_confusion_data_frame, save=True)
return rate_class, first_class