def learn(trace_fpath,
nk,
rate,
regularization,
alpha,
tau,
from_=0,
to=np.inf,
validation=0.1):
dts, Trace, seen, hyper2id, obj2id = \
dataio.initialize_trace(trace_fpath, from_, to)
no = len(obj2id)
nh = len(hyper2id)
validation_from = int(len(dts) - len(dts) * validation)
print('Using first %d of %d as train, rest is validation' \
% (validation_from, len(dts)))
dts_train = dts[:validation_from]
Trace_train = Trace[:validation_from]
rnd_idx = np.arange(len(dts_train))
np.random.shuffle(rnd_idx)
dts_train = np.asanyarray(dts_train[rnd_idx], dtype='f8', order='C')
Trace_train = np.asanyarray(Trace_train[rnd_idx], dtype='i4', order='C')
dts_val = np.asanyarray(dts[validation_from:], dtype='f8', order='C')
Trace_val = np.asanyarray(Trace[validation_from:], dtype='i4', order='C')
XG_ok = np.random.normal(0, 0.01, (no, nk))
XP_ok = np.random.normal(0, 0.01, (no, nk))
XP_hk = np.random.normal(0, 0.01, (nh, nk))
cost_train, cost_val = sgd(dts, Trace, XG_ok, XP_ok, XP_hk, seen, rate, \
regularization, alpha, tau, dts_val, Trace_val)
rv = {}
rv['num_topics'] = np.asarray([nk])
rv['trace_fpath'] = np.asarray([os.path.abspath(trace_fpath)])
rv['rate'] = np.asarray([rate])
rv['regularization'] = np.asarray([regularization])
rv['alpha'] = np.asarray([alpha])
rv['tau'] = np.asarray([tau])
rv['from_'] = np.asarray([from_])
rv['cost_train'] = np.asarray([cost_train])
rv['cost_val'] = np.asarray([cost_val])
rv['to'] = np.asarray([to])
rv['hyper2id'] = hyper2id
rv['obj2id'] = obj2id
rv['XG_ok'] = XG_ok
rv['XP_ok'] = XP_ok
rv['XP_hk'] = XP_hk
return rv
def learn(trace_fpath, nk, rate, regularization, alpha, tau, from_=0, to=np.inf, validation=0.1):
dts, Trace, seen, hyper2id, obj2id = dataio.initialize_trace(trace_fpath, from_, to)
no = len(obj2id)
nh = len(hyper2id)
validation_from = int(len(dts) - len(dts) * validation)
print("Using first %d of %d as train, rest is validation" % (validation_from, len(dts)))
dts_train = dts[:validation_from]
Trace_train = Trace[:validation_from]
rnd_idx = np.arange(len(dts_train))
np.random.shuffle(rnd_idx)
dts_train = np.asanyarray(dts_train[rnd_idx], dtype="f8", order="C")
Trace_train = np.asanyarray(Trace_train[rnd_idx], dtype="i4", order="C")
dts_val = np.asanyarray(dts[validation_from:], dtype="f8", order="C")
Trace_val = np.asanyarray(Trace[validation_from:], dtype="i4", order="C")
XG_ok = np.random.normal(0, 0.01, (no, nk))
XP_ok = np.random.normal(0, 0.01, (no, nk))
XP_hk = np.random.normal(0, 0.01, (nh, nk))
cost_train, cost_val = sgd(
dts, Trace, XG_ok, XP_ok, XP_hk, seen, rate, regularization, alpha, tau, dts_val, Trace_val
)
rv = {}
rv["num_topics"] = np.asarray([nk])
rv["trace_fpath"] = np.asarray([os.path.abspath(trace_fpath)])
rv["rate"] = np.asarray([rate])
rv["regularization"] = np.asarray([regularization])
rv["alpha"] = np.asarray([alpha])
rv["tau"] = np.asarray([tau])
rv["from_"] = np.asarray([from_])
rv["cost_train"] = np.asarray([cost_train])
rv["cost_val"] = np.asarray([cost_val])
rv["to"] = np.asarray([to])
rv["hyper2id"] = hyper2id
rv["obj2id"] = obj2id
rv["XG_ok"] = XG_ok
rv["XP_ok"] = XP_ok
rv["XP_hk"] = XP_hk
return rv
def learn(trace_fpath, nk, rate, regularization, sigma, \
from_=0, to=np.inf, validation=0.1):
_, Trace, seen, hyper2id, obj2id = \
dataio.initialize_trace(trace_fpath, from_, to)
no = len(obj2id)
nh = len(hyper2id)
validation_from = int(len(Trace) - len(Trace) * validation)
print('Using first %d of %d as train, rest is validation' \
% (validation_from, len(Trace)))
Trace_train = Trace[:validation_from]
rnd_idx = np.arange(len(Trace_train))
np.random.shuffle(rnd_idx)
Trace_train = np.asanyarray(Trace_train[rnd_idx], dtype='i4', order='C')
Trace_val = np.asanyarray(Trace[validation_from:], dtype='i4', order='C')
VUI_uk = np.random.normal(0, sigma, (nh, nk))
VIU_ok = np.random.normal(0, sigma, (no, nk))
VLI_ok = np.random.normal(0, sigma, (no, nk))
VIL_ok = np.random.normal(0, sigma, (no, nk))
cost_train, cost_val = sgd(Trace, VUI_uk, VIU_ok, VLI_ok, VIL_ok, \
rate, regularization, Trace_val)
rv = {}
rv['num_topics'] = np.asarray([nk])
rv['trace_fpath'] = np.asarray([os.path.abspath(trace_fpath)])
rv['rate'] = np.asarray([rate])
rv['regularization'] = np.asarray([regularization])
rv['from_'] = np.asarray([from_])
rv['to'] = np.asarray([to])
rv['hyper2id'] = hyper2id
rv['obj2id'] = obj2id
rv['cost_train'] = np.asarray([cost_train])
rv['cost_val'] = np.asarray([cost_val])
rv['VUI_uk'] = VUI_uk
rv['VIU_ok'] = VIU_ok
rv['VLI_ok'] = VLI_ok
rv['VIL_ok'] = VIL_ok
return rv
def learn(trace_fpath, nk, rate, regularization, sigma, \
from_=0, to=np.inf, validation=0.1):
_, Trace, seen, hyper2id, obj2id = \
dataio.initialize_trace(trace_fpath, from_, to)
no = len(obj2id)
nh = len(hyper2id)
validation_from = int(len(Trace) - len(Trace) * validation)
print('Using first %d of %d as train, rest is validation' \
% (validation_from, len(Trace)))
Trace_train = Trace[:validation_from]
rnd_idx = np.arange(len(Trace_train))
np.random.shuffle(rnd_idx)
Trace_train = np.asanyarray(Trace_train[rnd_idx], dtype='i4', order='C')
Trace_val = np.asanyarray(Trace[validation_from:], dtype='i4', order='C')
VUI_uk = np.random.normal(0, sigma, (nh, nk))
VIU_ok = np.random.normal(0, sigma, (no, nk))
VLI_ok = np.random.normal(0, sigma, (no, nk))
VIL_ok = np.random.normal(0, sigma, (no, nk))
cost_train, cost_val = sgd(Trace, VUI_uk, VIU_ok, VLI_ok, VIL_ok, \
rate, regularization, Trace_val)
rv = {}
rv['num_topics'] = np.asarray([nk])
rv['trace_fpath'] = np.asarray([os.path.abspath(trace_fpath)])
rv['rate'] = np.asarray([rate])
rv['regularization'] = np.asarray([regularization])
rv['from_'] = np.asarray([from_])
rv['to'] = np.asarray([to])
rv['hyper2id'] = hyper2id
rv['obj2id'] = obj2id
rv['cost_train'] = np.asarray([cost_train])
rv['cost_val'] = np.asarray([cost_val])
rv['VUI_uk'] = VUI_uk
rv['VIU_ok'] = VIU_ok
rv['VLI_ok'] = VLI_ok
rv['VIL_ok'] = VIL_ok
return rv
