def onpress(self, event):
if event.key == 'd':
mtgp = MultiOutputGP(n_channels=3, n_latent_gps=2)
t1 = time.time()
raw_ts = [zip(self.x, self.y)]
train_ts = mtgp.gen_collection(raw_ts)
mtgp.train(train_ts, maxiter=50)
t2 = time.time()
print 'time:', t2 - t1
#test_x = [np.linspace(self.min_x, self.max_x, 100)] * self.n_ts
test_x = np.linspace(self.min_x, self.max_x, 100)
#test_x = [[1, 2], [2, 3]]
#post_mean, post_cov = mtgp.predict(test_x)
post_mean, post_cov = mtgp.predictive_gaussian(train_ts[0], test_x)
for i in xrange(self.n_ts):
#std = np.sqrt(np.diag(post_cov[i]))
std = np.sqrt(np.diag(post_cov[i]))
ax = self.ax[i]
self.clear_ax(ax)
ax.fill_between(test_x,
post_mean[i] - std,
post_mean[i] + std,
edgecolor='none', alpha=.3, color='g')
ax.plot(test_x, post_mean[i], 'g-')
ax.plot(self.x[i], self.y[i], 'ko')
self.fig.canvas.draw()
elif event.key == 'm':
pickle_save('input.pkl', self.x, self.y)
elif event.key == 'c':
self.reset_data()
for ax in self.ax:
self.clear_ax(ax)
self.fig.canvas.draw()
def save_params(self, params_file):
# TODO: save t_test instead of idx_test and w_test
model_params = (self.net_arch, self.n_classes, self.inducing_pts,
self.idx_test, self.w_test, self.gp_output_len)
network_params = get_all_param_values(self.train_network)
gp_params = [p.get_value() for p in self.post_gp.params]
pickle_save(params_file, model_params, network_params, gp_params)
def main():
for sparsity in [5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100]:
#data = 'data/UWaveGestureLibraryAll-%d.pkl' % sparsity
data = 'data/PhalangesOutlinesCorrect-%d.pkl' % sparsity
print data
t1 = time.time()
gp_params, indep_noise = train_gp(data)
t2 = time.time()
print 'time', t2 - t1
data_id = data.rsplit('/', 1)[-1]
pickle_save('params/%s' % data_id, gp_params, indep_noise)
def cmp_n_train_test(save_as=None, n_inducing_pts=512):
np.random.seed(0)
batch_size = 1
results = []
for n_data in xrange(500, 3000 + 1, 500):
#n_data = 1000
#n_data = 1000
gp_params = np.array([.1, 100])
indep_noise = .001
x, y = gen_data(n_data, gp_params, indep_noise)
x_min, x_max = x.min(), x.max()
#len_u = 2048 + 1
print '# data', n_data
n_test = n_data
x_test = np.linspace(x_min, x_max, n_test)
proj_1d = np.random.uniform(-1, 1, size=len(x_test))
proj_1d /= np.linalg.norm(proj_1d)
proj_2d = np.random.uniform(-1, 1, size=(batch_size, len(x_test)))
proj_2d /= np.linalg.norm(proj_2d, axis=1)[:, np.newaxis]
t_exact = time_exact(x, y, x_test, gp_params, indep_noise, proj_1d,
proj_2d)
# parameters for approximation algorithm
len_u = n_inducing_pts
proj_1d = np.random.uniform(-1, 1, size=len(x_test))
proj_1d /= np.linalg.norm(proj_1d)
proj_2d = np.random.uniform(-1, 1, size=(batch_size, len(x_test)))
proj_2d /= np.linalg.norm(proj_2d, axis=1)[:, np.newaxis]
t_approx = time_approx(x, y, x_test, gp_params, indep_noise, len_u,
proj_1d, proj_2d)
print
results.append(np.concatenate((t_exact, t_approx)))
if save_as:
pickle_save(save_as, np.vstack(results))
def save_model(self, pickle_name):
pickle_save(pickle_name, self.gp_parms)
def main():
np.random.seed(0)
dat_id = 1
ts_all, l_all = pickle_load('../chla-data/chla_ts_min0_%d.pkl' % dat_id)
# randomly shuffle training examples
idx = np.arange(len(ts_all))
np.random.shuffle(idx)
ts_all = ts_all[idx]
#l_all = l_all[idx]
parser = argparse.ArgumentParser()
parser.add_argument('-q', dest='n_latent_gp', type=int, default=5)
parser.add_argument('-g', dest='w_reg_group', default='none')
parser.add_argument('-r', dest='w_reg', type=float, default=1)
args = parser.parse_args()
Q = args.n_latent_gp
w_reg_group = args.w_reg_group
w_reg = args.w_reg
P = len(ts_all[0])
#Q = 10
M = 20
#mogp = MultiOutputGP(P, Q, M)
#mogp = MultiOutputGP(P, Q, M, w_reg_group='individual', w_reg=.5)
#mogp = MultiOutputGP(P, Q, M, w_reg_group='row', w_reg=2)
#mogp = MultiOutputGP(P, Q, M, w_reg_group='column', w_reg=1)
#mogp = MultiOutputGP(P, Q, M, w_reg_group='column', w_reg=1.5)
mogp = MultiOutputGP(P, Q, M, w_reg_group=w_reg_group, w_reg=w_reg)
n_train = int(len(ts_all) * 0.5)
#n_train = 8
print 'n_train', n_train
train_raw = ts_all[:n_train]
train_ts = mogp.gen_collection(train_raw)
mogp.train(train_ts, maxiter=50)
w_reg_group_name = {
'none': 'non',
'row': 'row',
'column': 'col',
'individual': 'ind',
}
mogp_str = 'model-pqn-%s-%g-%d-%d' % (
w_reg_group_name[mogp.w_reg_group], mogp.w_reg, Q, n_train)
print mogp_str
mogp_pickle = 'model/%s.pkl' % mogp_str
mogp.load_model(mogp_pickle)
test_raw = ts_all[n_train:]
indep_gp = IndependentMultiOutputGP(P)
indep_gp.train(train_raw)
gp_parms = indep_gp.gp_parms
for channel in xrange(P):
loglike = []
loglike_baseline = []
for each_test in test_raw:
x = [xy[0] for xy in each_test]
y = [xy[1] for xy in each_test]
channel_len = len(x[channel])
if channel_len < 3:
continue
one_third = channel_len // 3
x_held_out = x[channel][one_third:-one_third]
y_held_out = y[channel][one_third:-one_third]
x_remain = [each_x if i == channel else
np.concatenate((each_x[:one_third],
each_x[-one_third:]))
for i, each_x in enumerate(x)]
y_remain = [each_y if i == channel else
np.concatenate((each_y[:one_third],
each_y[-one_third:]))
for i, each_y in enumerate(y)]
ts = TimeSeries(x_remain, y_remain, mogp.shared)
mu, cov = mogp.predictive_gaussian(ts, x_held_out)
mean, var = gp.pointwise_posterior_mean_var(
x_remain[channel], y_remain[channel], x_held_out,
gp_parms[channel])
for i, each_y in enumerate(y_held_out):
loglike.append(norm.logpdf(each_y, mu[channel, i],
np.sqrt(cov[channel, i, i])))
loglike_baseline.append(norm.logpdf(each_y, mean[i],
np.sqrt(var[i])))
print '%2d %10.2f %10.2f %10.2f %10.2f %6d' % (
channel, np.mean(loglike),
np.std(loglike) / np.sqrt(len(loglike)),
np.min(loglike), np.max(loglike), len(loglike))
print '%2d %10.2f %10.2f %10.2f %10.2f %6d' % (
channel, np.mean(loglike_baseline),
np.std(loglike_baseline) / np.sqrt(len(loglike_baseline)),
np.min(loglike_baseline), np.max(loglike_baseline),
len(loglike_baseline))
print '-' * 50
pickle_save('loglike-cmp/%02d.pkl' % channel, loglike, loglike_baseline)
pl.figure()
pl.axis('equal')
pl.scatter(loglike, loglike_baseline, alpha=.5)
pl.savefig('loglike-cmp/loglike-%02d.pdf' % channel)
pl.close()
def save_model(self, pickle_name):
shared = self.shared
pickle_save(pickle_name,
shared.w, shared.beta, shared.g_gp_b,
shared.h_gp_a, shared.h_gp_b)
def save_model(self, pickle_name):
pickle_save(pickle_name, self.gp_parms)
def main():
np.random.seed(0)
dat_id = 1
ts_all, l_all = pickle_load('../chla-data/chla_ts_min0_%d.pkl' % dat_id)
# randomly shuffle training examples
idx = np.arange(len(ts_all))
np.random.shuffle(idx)
ts_all = ts_all[idx]
#l_all = l_all[idx]
parser = argparse.ArgumentParser()
parser.add_argument('-q', dest='n_latent_gp', type=int, default=5)
parser.add_argument('-g', dest='w_reg_group', default='none')
parser.add_argument('-r', dest='w_reg', type=float, default=1)
args = parser.parse_args()
Q = args.n_latent_gp
w_reg_group = args.w_reg_group
w_reg = args.w_reg
P = len(ts_all[0])
#Q = 10
M = 20
#mogp = MultiOutputGP(P, Q, M)
#mogp = MultiOutputGP(P, Q, M, w_reg_group='individual', w_reg=.5)
#mogp = MultiOutputGP(P, Q, M, w_reg_group='row', w_reg=2)
#mogp = MultiOutputGP(P, Q, M, w_reg_group='column', w_reg=1)
#mogp = MultiOutputGP(P, Q, M, w_reg_group='column', w_reg=1.5)
mogp = MultiOutputGP(P, Q, M, w_reg_group=w_reg_group, w_reg=w_reg)
n_train = int(len(ts_all) * 0.5)
#n_train = 8
print 'n_train', n_train
train_raw = ts_all[:n_train]
train_ts = mogp.gen_collection(train_raw)
mogp.train(train_ts, maxiter=50)
w_reg_group_name = {
'none': 'non',
'row': 'row',
'column': 'col',
'individual': 'ind',
}
mogp_str = 'model-pqn-%s-%g-%d-%d' % (w_reg_group_name[mogp.w_reg_group],
mogp.w_reg, Q, n_train)
print mogp_str
mogp_pickle = 'model/%s.pkl' % mogp_str
mogp.load_model(mogp_pickle)
test_raw = ts_all[n_train:]
indep_gp = IndependentMultiOutputGP(P)
indep_gp.train(train_raw)
gp_parms = indep_gp.gp_parms
for channel in xrange(P):
loglike = []
loglike_baseline = []
for each_test in test_raw:
x = [xy[0] for xy in each_test]
y = [xy[1] for xy in each_test]
channel_len = len(x[channel])
if channel_len < 3:
continue
one_third = channel_len // 3
x_held_out = x[channel][one_third:-one_third]
y_held_out = y[channel][one_third:-one_third]
x_remain = [
each_x if i == channel else np.concatenate(
(each_x[:one_third], each_x[-one_third:]))
for i, each_x in enumerate(x)
]
y_remain = [
each_y if i == channel else np.concatenate(
(each_y[:one_third], each_y[-one_third:]))
for i, each_y in enumerate(y)
]
ts = TimeSeries(x_remain, y_remain, mogp.shared)
mu, cov = mogp.predictive_gaussian(ts, x_held_out)
mean, var = gp.pointwise_posterior_mean_var(
x_remain[channel], y_remain[channel], x_held_out,
gp_parms[channel])
for i, each_y in enumerate(y_held_out):
loglike.append(
norm.logpdf(each_y, mu[channel, i],
np.sqrt(cov[channel, i, i])))
loglike_baseline.append(
norm.logpdf(each_y, mean[i], np.sqrt(var[i])))
print '%2d %10.2f %10.2f %10.2f %10.2f %6d' % (
channel, np.mean(loglike), np.std(loglike) / np.sqrt(len(loglike)),
np.min(loglike), np.max(loglike), len(loglike))
print '%2d %10.2f %10.2f %10.2f %10.2f %6d' % (
channel, np.mean(loglike_baseline), np.std(loglike_baseline) /
np.sqrt(len(loglike_baseline)), np.min(loglike_baseline),
np.max(loglike_baseline), len(loglike_baseline))
print '-' * 50
pickle_save('loglike-cmp/%02d.pkl' % channel, loglike,
loglike_baseline)
pl.figure()
pl.axis('equal')
pl.scatter(loglike, loglike_baseline, alpha=.5)
pl.savefig('loglike-cmp/loglike-%02d.pdf' % channel)
pl.close()
def save_model(self, pickle_name):
shared = self.shared
pickle_save(pickle_name, shared.w, shared.beta, shared.g_gp_b,
shared.h_gp_a, shared.h_gp_b)
t2 = time.time()
print t2 - t1
print
#print '-' * 40
#print (cov_zs**2).sum(axis=0)
#print (cov_zs_exact**2).sum(axis=0)
if __name__ == '__main__':
np.random.seed(0)
#cmp_time()
#errors = cmp_lanczos_basis(500)
for i in xrange(10):
results = []
for n_data in xrange(500, 3000 + 1, 500):
errors = cmp_lanczos_basis(n_data)
results.append(errors)
pickle_save('results/lanczos-%02d.pkl' % i, np.vstack(results))
'''
for i in xrange(10):
results = []
for n_data in xrange(200, 3000 + 1, 200):
errors = cmp_n_inducing_points(n_data)
results.append(errors)
pickle_save('results/error-%02d.pkl' % i, np.vstack(results))
'''
#norm_diff()
if __name__ == '__main__':
descriptions = np.array(
pd.read_csv("../data/descriptions.csv")['description'])
scores = np.array(pd.read_csv("../data/scores.csv")['points'])
tfidf = TfidfVectorizer(ngram_range=(1, 2))
X = tfidf.fit_transform(descriptions)
X_train, X_test, y_train, y_test = train_test_split(X,
scores,
test_size=0.2,
random_state=42)
rf_regression = RandomForestRegressor().fit(X_train, y_train)
y_train_pred = rf_regression.predict(X_train)
y_test_pred = rf_regression.predict(X_test)
print metrics.mean_squared_error(y_train, y_train_pred)
# 0.506535834654
print metrics.mean_squared_error(y_test, y_test_pred)
# 2.79865706702
#
pickle_io.pickle_save("../models/random_forest_model.pkl", rf_regression)
pickle_io.pickle_save("../vectorizers/random_forest_vectorizer.pkl", tfidf)
# wine_regressor = load_regressor()
