NEPOCH = 50
APOSTLE = 'v4_nopt'
system('cp %s shallow_models/train_%s.py' % (argv[0], APOSTLE))
'''
instantiate data loaders
'''
top = make_coll(basedir + '/PARTITION/Top_*_CATEGORY.npy')
qcd = make_coll(basedir + '/PARTITION/QCD_*_CATEGORY.npy')
data = [top, qcd]
'''
first build the classifier!
'''
# set up data
classifier_train_gen = generate(data, partition='train', batch=1000)
classifier_validation_gen = generate(data, partition='validate', batch=10000)
classifier_test_gen = generate(data, partition='test', batch=10)
test_i, test_o, test_w = next(classifier_test_gen)
#print test_i
inputs = Input(shape=(len(config.gen_default_variables), ), name='input')
dense = Dense(32,
activation='tanh',
name='dense1',
kernel_initializer='lecun_uniform')(inputs)
dense = Dense(32,
activation='tanh',
name='dense2',
kernel_initializer='lecun_uniform')(dense)
dense = Dense(32,
APOSTLE = 'v4_0'
modeldir = 'cce_adversary/'
system('mkdir -p %s'%modeldir)
system('cp %s %s/train_%s.py'%(argv[0], modeldir, APOSTLE))
### instantiate data loaders ###
top = make_coll(basedir + '/PARTITION/Top_*_CATEGORY.npy')
qcd = make_coll(basedir + '/PARTITION/QCD_*_CATEGORY.npy')
data = [top, qcd]
### first build the classifier! ###
# set up data
opts = {'decorr_mass':False}
classifier_train_gen = generate(data, partition='train', batch=1000, **opts)
classifier_validation_gen = generate(data, partition='validate', batch=10000, **opts)
classifier_test_gen = generate(data, partition='test', batch=10, **opts)
test_i, test_o, test_w = next(classifier_test_gen)
inputs = Input(shape=(len(config.gen_default_variables),), name='input')
dense = Dense(32, activation='tanh',name='dense1',kernel_initializer='lecun_uniform') (inputs)
dense = Dense(32, activation='tanh',name='dense2',kernel_initializer='lecun_uniform') (dense)
dense = Dense(32, activation='tanh',name='dense3',kernel_initializer='lecun_uniform') (dense)
y_hat = Dense(config.n_truth, activation='softmax') (dense)
classifier = Model(inputs=[inputs], outputs=[y_hat])
classifier.compile(optimizer=Adam(lr=0.0005),
loss=['categorical_crossentropy'],
metrics=['accuracy'])