jitonic = None
tau_cla = 50000
namelist = ['raw', 'homhots']
namelist = ['raw', 'hots', 'homhots']
for name in namelist:
f_name = f'{record_path}{timestr}_LR_results_{name}_{nbclust}_36_40_{ds_ev}.pkl'
if os.path.isfile(f_name):
with open(f_name, 'rb') as file:
likelihood, true_target, time_scale = pickle.load(file)
else:
print(f'LR fit for {name}...')
model, loss = fit_data(name,timestr,record_path,filt,tau,R,nbclust,sigma,homeinv,jitter,dataset,nb_train, ds_ev,learning_rate,num_epochs,betas,tau_cla,jitonic=jitonic,num_workers=num_workers,verbose=False)
print(f'prediction for {name}...')
likelihood, true_target, time_scale = predict_data(model,name,timestr,record_path,filt,tau,R,nbclust,sigma, homeinv, jitter,dataset,nb_test,ds_ev,tau_cla,jitonic=jitonic,num_workers=num_workers, verbose=False)
with open(f_name, 'wb') as file:
pickle.dump([likelihood, true_target, time_scale], file, pickle.HIGHEST_PROTOCOL)
timesteps = np.arange(500,100000,100)
nb_classes = 10
results = [timesteps]
for namnum, name in enumerate(namelist):
f_name = f'{record_path}{timestr}_LR_results_{name}_{nbclust}_{nb_train}_{nb_test}_{ds_ev}_timescale.pkl'
with open(f_name, 'rb') as file:
likelihood, true_target, time_scale = pickle.load(file)
y_true = np.zeros([len(true_target)])
beta1, beta2 = 0.9, 0.999
betas = (beta1, beta2)
num_epochs = 2 ** 5 + 1
#num_epochs = 2 ** 9 + 1
print(f'number of epochs: {num_epochs}')
#______________________________________________
timestr = '2021-02-16'
record_path = '../Records/EXP_03_NMNIST/models/'
for name in ['raw','hots','homhots']:
tic()
learn_set, nb_pola, name_net = get_loader(name, record_path, nb_train, True, filt, tau, nblay, nbclust, sigma, homeinv, jitter, timestr)
model, loss = fit_data(name_net, learn_set, nb_train, nb_pola, learning_rate, num_epochs, betas, verbose=True)
test_set, nb_pola, name_net = get_loader(name, record_path, nb_test, False, filt, tau, nblay, nbclust, sigma, homeinv, jitter, timestr)
pred_target, true_target = predict_data(test_set, model, nb_test)
mean_acc, online_acc = classification_results(pred_target, true_target, nb_test)
toc()
print(f'Classification performance for {name}: {mean_acc}')
homhots_jit_s = []
hots_jit_s = []
homhots_jit_t = []
hots_jit_t = []
for name in ['homhots', 'hots']:
learn_set, nb_pola, name_net = get_loader(name, record_path, nb_train, True, filt, tau, nblay, nbclust, sigma, homeinv, jitter, timestr)
model, loss = fit_data(name_net, learn_set, nb_train, nb_pola, learning_rate, num_epochs, betas, verbose=True)
for j_s in jit_s:
jitonic = [None,j_s]
test_set, nb_pola, name_net = get_loader(name, record_path, nb_test, False, filt, tau, nblay, nbclust, sigma, homeinv, jitter, timestr, jitonic = jitonic)
pred_target, true_target = predict_data(test_set, model, nb_test)
print(f'get testing set for {name}...')
test_set, nb_pola, name_net = get_loader(name,
record_path,
nb_test,
False,
filt,
tau,
nbclust,
sigma,
homeinv,
jitter,
timestr,
dataset,
R,
ds_ev=ds_ev_output)
print(f'prediction for {name}...')
pred_target, true_target = predict_data(test_set,
model,
nb_test,
num_workers=num_workers)
mean_acc, online_acc = classification_results(pred_target, true_target,
nb_test)
print(f'Classification performance for {name}: {mean_acc}')
results.append([pred_target, true_target, mean_acc, online_acc])
path = '../Records/EXP_05_POKERDVS/'
f_name = f'{path}{timestr}_LR_results_{nbclust}_{nb_train}_{nb_test}_{ds_ev_output}.pkl'
with open(f_name, 'wb') as file:
pickle.dump([results], file, pickle.HIGHEST_PROTOCOL)
jitonic=jitonic,
subset_size=nb_train,
num_workers=num_workers,
verbose=False)
ds_ev = ds_ev_test
print(f'prediction for {name}...')
likelihood, true_target, time_scale = predict_data(
model,
name,
timestr,
path,
filt,
tau,
R,
nbclust,
sigma,
homeinv,
jitter,
dataset,
nb_test,
ds_ev,
tau_cla,
sample_space=sample_space,
jitonic=jitonic,
subset_size=nb_test,
num_workers=num_workers,
verbose=False)
with open(f_name, 'wb') as file:
pickle.dump([likelihood, true_target, time_scale], file,
pickle.HIGHEST_PROTOCOL)
#return likelihood, true_target, time_scale