print 'sumcut:', sumcut, 'ptcut:', ptcut, 'mrcut:', mrcut, 'r2cut:', r2cut
dataset = 'SOM%s_sumMET%d_metPt%d_MR%d_Rsq%d.hdf5'%(label, sumcut, ptcut, mrcut, r2cut * 100)
print dataset
mappath = os.path.join(datapath, 'SOM', dataset[0:-5])
datafile = os.path.join(datapath, dataset)
logfile = os.path.join(mappath, 'results.log')
log = open(logfile, 'w')
exclusion = []
discovery = []
for mapfile in [os.path.join(mappath, f) for f in os.listdir(mappath) if f.endswith('hdf5')]:
print mapfile
Map = SOM(datafile, shape = (10, 10, 10), filename = mapfile)
log.write('Epochs: {0:2d}, Sigma: ({1:4f}, {2:4f}), Lrate: ({3:4f}, {4:4f}) \n'.format(Map.epochs, Map.sigma_i, Map.sigma_f, Map.lrate_i, Map.lrate_f))
Map.test_theano(quiet = True)
Map.show_map()
_, zeros = Map.negloglikelihood([3])
log.write('contains {0:2d} zeros \n'.format(zeros))
exc = Map.exclusion((0, 2), n = 1000, show = True)
dis = Map.discovery((0, 3), n = 1000, show = True)
log.write('exclusion: {0:5f} \n'.format(exc))
log.write('discovery: {0:5f} \n'.format(dis))
print 'contains {0:2d} zeros'.format(zeros)
print exc, dis
exclusion.append(exc)
discovery.append(dis)
mean_exc = np.mean(exclusion)
resultspath = os.path.join(datapath, 'SOM', dataset[0:-5])
if not os.path.exists(resultspath):
os.mkdir(resultspath)
logfile = os.path.join(resultspath, 'training.log')
log = open(logfile, 'a')
log.write('---- Training with no razor variables ---- \n')
for e in epochs:
for s in sigma:
for l in lrate:
for i in range(trials):
log.write(
'Epochs: {0:2d}, Sigma: ({1:4f}, {2:4f}), Lrate: ({3:4f}, {4:4f}) \n'
.format(e, s[0], s[1], l[0], l[1]))
Map = SOM(datafile, razor=False, shape=(10, 10, 10))
Map.set_params(epochs=e,
lrate=l,
sigma=s,
threshold=threshold)
Map.train_theano(quiet=True)
Map.test_theano(quiet=True)
# Map.show_map()
_, zeros = Map.negloglikelihood([3])
exc = Map.exclusion((0, 5), n=1000, show=False)
dis = Map.discovery((0, 5), n=1000, show=False)
log.write('contains {0:2d} zeros \n'.format(zeros))
log.write('exclusion: {0:5f} \n'.format(exc))
log.write('discovery: {0:5f} \n \n'.format(dis))
print 'contains {0:2d} zeros'.format(zeros)
print 'exclusion: {0:5f} '.format(exc)
datafile = os.path.join(datapath, dataset)
resultspath = os.path.join(datapath, 'SOM', dataset[0:-5])
if not os.path.exists(resultspath):
os.mkdir(resultspath)
logfile = os.path.join(resultspath, 'training.log')
log = open(logfile, 'a')
log.write('---- Training with no razor variables ---- \n')
for e in epochs:
for s in sigma:
for l in lrate:
for i in range(trials):
log.write('Epochs: {0:2d}, Sigma: ({1:4f}, {2:4f}), Lrate: ({3:4f}, {4:4f}) \n'.format(e, s[0], s[1], l[0], l[1]))
Map = SOM(datafile, razor = False, shape = (10, 10, 10))
Map.set_params(epochs = e, lrate = l, sigma = s, threshold = threshold)
Map.train_theano(quiet = True)
Map.test_theano(quiet = True)
# Map.show_map()
_, zeros = Map.negloglikelihood([3])
exc = Map.exclusion((0, 5), n = 1000, show = False)
dis = Map.discovery((0, 5), n = 1000, show = False)
log.write('contains {0:2d} zeros \n'.format(zeros))
log.write('exclusion: {0:5f} \n'.format(exc))
log.write('discovery: {0:5f} \n \n'.format(dis))
print 'contains {0:2d} zeros'.format(zeros)
print 'exclusion: {0:5f} '.format(exc)
print 'discovery: {0:5f} '.format(dis)
# Map.save_map(resultspath)
log.close()
