def __init__(self, params: LogisticParam):
super(HomoLRHost, self).__init__(params)
self.learning_rate = params.learning_rate
self.batch_size = params.batch_size
self.encrypt_params = params.encrypt_param
if self.encrypt_params.method in [consts.PAILLIER]:
self.use_encrypt = True
else:
self.use_encrypt = False
if self.use_encrypt and params.penalty != consts.L2_PENALTY:
raise RuntimeError("Encrypted h**o-lr supports L2 penalty only")
if self.use_encrypt:
self.gradient_operator = TaylorLogisticGradient()
self.re_encrypt_batches = params.re_encrypt_batches
else:
self.gradient_operator = LogisticGradient()
self.aggregator = HomoFederatedAggregator()
self.party_weight = params.party_weight
self.optimizer = Optimizer(learning_rate=self.learning_rate, opt_method_name=params.optimizer)
self.transfer_variable = HomoLRTransferVariable()
self.initializer = Initializer()
self.mini_batch_obj = None
self.evaluator = Evaluation(classi_type=consts.BINARY)
self.classes_ = [0, 1]
self.has_sychronized_encryption = False
def __init__(self):
super(HomoLRHost, self).__init__()
self.aggregator = HomoFederatedAggregator()
self.initializer = Initializer()
self.mini_batch_obj = None
self.classes_ = [0, 1]
self.has_sychronized_encryption = False
self.role = consts.HOST
def test_initializer(self):
initializer = Initializer()
data_shape = 10
init_param_obj = InitParam(init_method=consts.RANDOM_NORMAL,
init_const=20,
fit_intercept=False)
model = initializer.init_model(model_shape=data_shape,
init_params=init_param_obj)
model_shape = model.shape
self.assertTrue(model_shape == (10, ))
def __init__(self):
super(HomoLRGuest, self).__init__()
self.aggregator = HomoFederatedAggregator
self.gradient_operator = LogisticGradient()
self.initializer = Initializer()
self.classes_ = [0, 1]
self.evaluator = Evaluation()
self.loss_history = []
self.is_converged = False
self.role = consts.GUEST
def __init__(self, logistic_params: LogisticParam):
# set params
self.alpha = logistic_params.alpha
self.init_param_obj = logistic_params.init_param
self.fit_intercept = self.init_param_obj.fit_intercept
self.learning_rate = logistic_params.learning_rate
if logistic_params.penalty == consts.L1_PENALTY:
self.updater = L1Updater(self.alpha, self.learning_rate)
elif logistic_params.penalty == consts.L2_PENALTY:
self.updater = L2Updater(self.alpha, self.learning_rate)
else:
self.updater = None
self.eps = logistic_params.eps
self.batch_size = logistic_params.batch_size
self.max_iter = logistic_params.max_iter
if logistic_params.encrypt_param.method == consts.PAILLIER:
self.encrypt_operator = PaillierEncrypt()
else:
self.encrypt_operator = FakeEncrypt()
# attribute:
self.n_iter_ = 0
self.coef_ = None
self.intercept_ = 0
self.classes_ = None
self.gradient_operator = None
self.initializer = Initializer()
self.transfer_variable = None
self.model_meta = LogisticRegressionModelMeta()
def __init__(self):
super(BaseLogisticRegression, self).__init__()
self.model_param = LogisticParam()
# attribute:
self.n_iter_ = 0
self.coef_ = None
self.intercept_ = 0
self.classes_ = None
self.feature_shape = None
self.gradient_operator = None
self.initializer = Initializer()
self.transfer_variable = None
self.loss_history = []
self.is_converged = False
self.header = None
self.class_name = self.__class__.__name__
self.model_name = 'LogisticRegression'
self.model_param_name = 'LogisticRegressionParam'
self.model_meta_name = 'LogisticRegressionMeta'
self.role = ''
self.mode = ''
self.schema = {}
# one_ve_rest parameter
self.need_one_vs_rest = False
self.one_vs_rest_classes = []
self.one_vs_rest_obj = None
def __init__(self, params: LogisticParam):
super(HomoLRGuest, self).__init__(params)
self.learning_rate = params.learning_rate
self.aggregator = HomoFederatedAggregator
self.gradient_operator = LogisticGradient()
self.party_weight = params.party_weight
self.optimizer = Optimizer(learning_rate=self.learning_rate,
opt_method_name=params.optimizer)
self.transfer_variable = HomoLRTransferVariable()
self.initializer = Initializer()
self.classes_ = [0, 1]
self.evaluator = Evaluation()
self.header = []
self.penalty = params.penalty
self.loss_history = []
self.is_converged = False
def setUp(self):
self.guest_X = np.array([[1, 2, 3, 4, 5], [3, 2, 4, 5, 1],
[
2,
2,
3,
1,
1,
]]) / 10
self.guest_Y = np.array([[1], [1], [-1]])
self.values = []
for idx, x in enumerate(self.guest_X):
inst = Instance(inst_id=idx, features=x, label=self.guest_Y[idx])
self.values.append((idx, inst))
self.host_X = np.array([[1, 1.2, 3.1, 4, 5], [2.3, 2, 4, 5.3, 1],
[
2,
2.2,
1.3,
1,
1.6,
]]) / 10
self.host_Y = np.array([[-1], [1], [-1]])
self.host_values = []
for idx, x in enumerate(self.host_X):
inst = Instance(inst_id=idx, features=x, label=self.host_Y[idx])
self.values.append((idx, inst))
self.max_iter = 10
self.alpha = 0.01
self.learning_rate = 0.01
optimizer = 'SGD'
self.gradient_operator = LogisticGradient()
self.initializer = Initializer()
self.fit_intercept = True
self.init_param_obj = InitParam(fit_intercept=self.fit_intercept)
self.updater = L2Updater(self.alpha, self.learning_rate)
self.optimizer = Optimizer(learning_rate=self.learning_rate,
opt_method_name=optimizer)
self.__init_model()
def __init__(self, params: LogisticParam):
"""
:param penalty: l1 or l2
:param alpha:
:param lr:
:param eps:
:param max_iter:
:param optim_method: must be in ['sgd', 'RMSProp' ,'Adam', 'AdaGrad']
:param batch_size: only work when otpim_method is mini-batch, represent for mini-batch's size
"""
super(HomoLRGuest, self).__init__(params)
self.learning_rate = params.learning_rate
self.aggregator = HomoFederatedAggregator
self.gradient_operator = LogisticGradient()
self.party_weight = params.party_weight
self.optimizer = Optimizer(learning_rate=self.learning_rate,
opt_method_name=params.optimizer)
self.transfer_variable = HomoLRTransferVariable()
self.initializer = Initializer()
self.classes_ = [0, 1]
self.evaluator = Evaluation()
def __init__(self, logistic_params: LogisticParam):
self.param = logistic_params
# set params
LogisticParamChecker.check_param(logistic_params)
self.alpha = logistic_params.alpha
self.init_param_obj = logistic_params.init_param
self.fit_intercept = self.init_param_obj.fit_intercept
self.learning_rate = logistic_params.learning_rate
self.encrypted_mode_calculator_param = logistic_params.encrypted_mode_calculator_param
self.encrypted_calculator = None
if logistic_params.penalty == consts.L1_PENALTY:
self.updater = L1Updater(self.alpha, self.learning_rate)
elif logistic_params.penalty == consts.L2_PENALTY:
self.updater = L2Updater(self.alpha, self.learning_rate)
else:
self.updater = None
self.eps = logistic_params.eps
self.batch_size = logistic_params.batch_size
self.max_iter = logistic_params.max_iter
if logistic_params.encrypt_param.method == consts.PAILLIER:
self.encrypt_operator = PaillierEncrypt()
else:
self.encrypt_operator = FakeEncrypt()
# attribute:
self.n_iter_ = 0
self.coef_ = None
self.intercept_ = 0
self.classes_ = None
self.feature_shape = None
self.gradient_operator = None
self.initializer = Initializer()
self.transfer_variable = None
self.model_meta = LogisticRegressionModelMeta()
self.loss_history = []
self.is_converged = False
self.header = None
self.class_name = self.__class__.__name__
class HomoLRHost(HomoLRBase):
def __init__(self):
super(HomoLRHost, self).__init__()
self.aggregator = HomoFederatedAggregator()
self.initializer = Initializer()
self.mini_batch_obj = None
self.classes_ = [0, 1]
self.has_sychronized_encryption = False
self.role = consts.HOST
def _init_model(self, params):
super(HomoLRHost, self)._init_model(params)
encrypt_params = params.encrypt_param
if encrypt_params.method in [consts.PAILLIER]:
self.use_encrypt = True
else:
self.use_encrypt = False
if self.use_encrypt and params.penalty == 'L1':
raise RuntimeError(
"Encrypted h**o-lr supports L2 penalty or 'none' only")
if self.use_encrypt:
self.gradient_operator = TaylorLogisticGradient()
self.re_encrypt_batches = params.re_encrypt_batches
else:
self.gradient_operator = LogisticGradient()
def fit(self, data_instances):
if not self.need_run:
return data_instances
self.init_schema(data_instances)
LOGGER.debug("Before trainning, self.header: {}".format(self.header))
self._abnormal_detection(data_instances)
self.__init_parameters(data_instances)
w = self.__init_model(data_instances)
for iter_num in range(self.max_iter):
# mini-batch
LOGGER.debug("In iter: {}".format(iter_num))
batch_data_generator = self.mini_batch_obj.mini_batch_data_generator(
)
batch_num = 0
total_loss = 0
for batch_data in batch_data_generator:
f = functools.partial(self.gradient_operator.compute,
coef=self.coef_,
intercept=self.intercept_,
fit_intercept=self.fit_intercept)
grad_loss = batch_data.mapPartitions(f)
n = batch_data.count()
if not self.use_encrypt:
grad, loss = grad_loss.reduce(
self.aggregator.aggregate_grad_loss)
grad = np.array(grad)
grad /= n
loss /= n
if self.updater is not None:
loss_norm = self.updater.loss_norm(self.coef_)
total_loss += loss + loss_norm
# if not self.use_loss:
# total_loss = np.linalg.norm(self.coef_)
if not self.need_one_vs_rest:
metric_meta = MetricMeta(
name='train',
metric_type="LOSS",
extra_metas={"unit_name": "iters"})
metric_name = self.get_metric_name('loss')
self.callback_meta(metric_name=metric_name,
metric_namespace='train',
metric_meta=metric_meta)
self.callback_metric(
metric_name=metric_name,
metric_namespace='train',
metric_data=[Metric(iter_num, total_loss)])
else:
grad, _ = grad_loss.reduce(self.aggregator.aggregate_grad)
grad = np.array(grad)
grad /= n
self.update_model(grad)
w = self.merge_model()
batch_num += 1
if self.use_encrypt and batch_num % self.re_encrypt_batches == 0:
to_encrypt_model_id = self.transfer_variable.generate_transferid(
self.transfer_variable.to_encrypt_model, iter_num,
batch_num)
federation.remote(
w,
name=self.transfer_variable.to_encrypt_model.name,
tag=to_encrypt_model_id,
role=consts.ARBITER,
idx=0)
re_encrypted_model_id = self.transfer_variable.generate_transferid(
self.transfer_variable.re_encrypted_model, iter_num,
batch_num)
LOGGER.debug("re_encrypted_model_id: {}".format(
re_encrypted_model_id))
w = federation.get(
name=self.transfer_variable.re_encrypted_model.name,
tag=re_encrypted_model_id,
idx=0)
w = np.array(w)
self.set_coef_(w)
model_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.host_model, iter_num)
federation.remote(w,
name=self.transfer_variable.host_model.name,
tag=model_transfer_id,
role=consts.ARBITER,
idx=0)
if not self.use_encrypt:
loss_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.host_loss, iter_num)
federation.remote(total_loss,
name=self.transfer_variable.host_loss.name,
tag=loss_transfer_id,
role=consts.ARBITER,
idx=0)
LOGGER.debug("model and loss sent")
final_model_id = self.transfer_variable.generate_transferid(
self.transfer_variable.final_model, iter_num)
w = federation.get(name=self.transfer_variable.final_model.name,
tag=final_model_id,
idx=0)
w = np.array(w)
self.set_coef_(w)
converge_flag_id = self.transfer_variable.generate_transferid(
self.transfer_variable.converge_flag, iter_num)
converge_flag = federation.get(
name=self.transfer_variable.converge_flag.name,
tag=converge_flag_id,
idx=0)
self.n_iter_ = iter_num
LOGGER.debug("converge_flag: {}".format(converge_flag))
if converge_flag:
break
# self.save_model()
def __init_parameters(self, data_instances):
party_weight_id = self.transfer_variable.generate_transferid(
self.transfer_variable.host_party_weight)
LOGGER.debug(
"Start to remote party_weight: {}, transfer_id: {}".format(
self.party_weight, party_weight_id))
federation.remote(self.party_weight,
name=self.transfer_variable.host_party_weight.name,
tag=party_weight_id,
role=consts.ARBITER,
idx=0)
self.__synchronize_encryption()
# Send re-encrypt times
self.mini_batch_obj = MiniBatch(data_inst=data_instances,
batch_size=self.batch_size)
if self.use_encrypt:
# LOGGER.debug("Use encryption, send re_encrypt_times")
total_batch_num = self.mini_batch_obj.batch_nums
re_encrypt_times = total_batch_num // self.re_encrypt_batches
transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.re_encrypt_times)
LOGGER.debug(
"Start to remote re_encrypt_times: {}, transfer_id: {}".format(
re_encrypt_times, transfer_id))
federation.remote(
re_encrypt_times,
name=self.transfer_variable.re_encrypt_times.name,
tag=transfer_id,
role=consts.ARBITER,
idx=0)
LOGGER.info("sent re_encrypt_times: {}".format(re_encrypt_times))
def __synchronize_encryption(self, mode='train'):
"""
Communicate with hosts. Specify whether use encryption or not and transfer the public keys.
"""
# Send if this host use encryption or not
use_encryption_id = self.transfer_variable.generate_transferid(
self.transfer_variable.use_encrypt, mode)
LOGGER.debug("Start to remote use_encrypt: {}, transfer_id: {}".format(
self.use_encrypt, use_encryption_id))
federation.remote(self.use_encrypt,
name=self.transfer_variable.use_encrypt.name,
tag=use_encryption_id,
role=consts.ARBITER,
idx=0)
# Set public key
if self.use_encrypt:
pubkey_id = self.transfer_variable.generate_transferid(
self.transfer_variable.paillier_pubkey, mode)
pubkey = federation.get(
name=self.transfer_variable.paillier_pubkey.name,
tag=pubkey_id,
idx=0)
LOGGER.debug("Received pubkey")
self.encrypt_operator.set_public_key(pubkey)
LOGGER.info("Finish synchronized ecryption")
self.has_sychronized_encryption = True
def predict(self, data_instances):
if not self.need_run:
return data_instances
if not self.has_sychronized_encryption:
self.__synchronize_encryption(mode='predict')
self.__load_arbiter_model()
else:
LOGGER.info("in predict, has synchronize encryption information")
feature_shape = get_features_shape(data_instances)
LOGGER.debug("Shape of coef_ : {}, feature shape: {}".format(
len(self.coef_), feature_shape))
local_data = data_instances.first()
LOGGER.debug("One data, features: {}".format(local_data[1].features))
wx = self.compute_wx(data_instances, self.coef_, self.intercept_)
if self.use_encrypt:
encrypted_wx_id = self.transfer_variable.generate_transferid(
self.transfer_variable.predict_wx)
LOGGER.debug("Host encrypted wx id: {}".format(encrypted_wx_id))
LOGGER.debug("Start to remote wx: {}, transfer_id: {}".format(
wx, encrypted_wx_id))
federation.remote(wx,
name=self.transfer_variable.predict_wx.name,
tag=encrypted_wx_id,
role=consts.ARBITER,
idx=0)
predict_result_id = self.transfer_variable.generate_transferid(
self.transfer_variable.predict_result)
LOGGER.debug("predict_result_id: {}".format(predict_result_id))
predict_result = federation.get(
name=self.transfer_variable.predict_result.name,
tag=predict_result_id,
idx=0)
# local_predict_table = predict_result.collect()
LOGGER.debug(
"predict_result count: {}, data_instances count: {}".format(
predict_result.count(), data_instances.count()))
predict_result_table = predict_result.join(
data_instances,
lambda p, d: [d.label, None, p, {
"0": None,
"1": None
}])
else:
pred_prob = wx.mapValues(lambda x: activation.sigmoid(x))
pred_label = self.classified(pred_prob,
self.predict_param.threshold)
if self.predict_param.with_proba:
predict_result = data_instances.mapValues(lambda x: x.label)
predict_result = predict_result.join(pred_prob, lambda x, y:
(x, y))
else:
predict_result = data_instances.mapValues(lambda x:
(x.label, None))
predict_result_table = predict_result.join(
pred_label,
lambda x, y: [x[0], y, x[1], {
"0": None,
"1": None
}])
LOGGER.debug("Finish predict")
LOGGER.debug("In host predict, predict_result_table is : {}".format(
predict_result_table.first()))
return predict_result_table
def __init_model(self, data_instances):
model_shape = data_overview.get_features_shape(data_instances)
w = self.initializer.init_model(model_shape,
init_params=self.init_param_obj)
w = self.encrypt_operator.encrypt_list(w)
w = np.array(w)
if self.fit_intercept:
self.coef_ = w[:-1]
self.intercept_ = w[-1]
else:
self.coef_ = w
self.intercept_ = 0
return w
def __load_arbiter_model(self):
final_model_id = self.transfer_variable.generate_transferid(
self.transfer_variable.final_model, "predict")
final_model = federation.get(
name=self.transfer_variable.final_model.name,
tag=final_model_id,
idx=0)
# LOGGER.info("Received arbiter's model")
# LOGGER.debug("final_model: {}".format(final_model))
self.set_coef_(final_model)
def _get_param(self):
if self.need_one_vs_rest:
one_vs_rest_class = list(map(str, self.one_vs_rest_obj.classes))
param_protobuf_obj = lr_model_param_pb2.LRModelParam(
iters=self.n_iter_,
loss_history=[],
is_converged=self.is_converged,
weight={},
intercept=0,
need_one_vs_rest=self.need_one_vs_rest,
one_vs_rest_classes=one_vs_rest_class)
return param_protobuf_obj
header = self.header
weight_dict = {}
for idx, header_name in enumerate(header):
coef_i = self.coef_[idx]
weight_dict[header_name] = coef_i
param_protobuf_obj = lr_model_param_pb2.LRModelParam(
iters=self.n_iter_,
loss_history=[],
is_converged=self.is_converged,
weight={},
intercept=0,
need_one_vs_rest=self.need_one_vs_rest,
header=header)
from google.protobuf import json_format
json_result = json_format.MessageToJson(param_protobuf_obj)
LOGGER.debug("json_result: {}".format(json_result))
return param_protobuf_obj
class TestHomoLR(unittest.TestCase):
def setUp(self):
self.guest_X = np.array([[1, 2, 3, 4, 5], [3, 2, 4, 5, 1],
[
2,
2,
3,
1,
1,
]]) / 10
self.guest_Y = np.array([[1], [1], [-1]])
self.values = []
for idx, x in enumerate(self.guest_X):
inst = Instance(inst_id=idx, features=x, label=self.guest_Y[idx])
self.values.append((idx, inst))
self.host_X = np.array([[1, 1.2, 3.1, 4, 5], [2.3, 2, 4, 5.3, 1],
[
2,
2.2,
1.3,
1,
1.6,
]]) / 10
self.host_Y = np.array([[-1], [1], [-1]])
self.host_values = []
for idx, x in enumerate(self.host_X):
inst = Instance(inst_id=idx, features=x, label=self.host_Y[idx])
self.values.append((idx, inst))
self.max_iter = 10
self.alpha = 0.01
self.learning_rate = 0.01
optimizer = 'SGD'
self.gradient_operator = LogisticGradient()
self.initializer = Initializer()
self.fit_intercept = True
self.init_param_obj = InitParam(fit_intercept=self.fit_intercept)
self.updater = L2Updater(self.alpha, self.learning_rate)
self.optimizer = Optimizer(learning_rate=self.learning_rate,
opt_method_name=optimizer)
self.__init_model()
def __init_model(self):
model_shape = self.guest_X.shape[1]
w = self.initializer.init_model(model_shape,
init_params=self.init_param_obj)
if self.fit_intercept:
self.coef_ = w[:-1]
self.intercept_ = w[-1]
else:
self.coef_ = w
self.intercept_ = 0
return w
def __init_host_model(self):
model_shape = self.host_X.shape[1]
w = self.initializer.init_model(model_shape,
init_params=self.init_param_obj)
if self.fit_intercept:
self.host_coef_ = w[:-1]
self.host_intercept_ = w[-1]
else:
self.host_coef_ = w
self.host_intercept_ = 0
return w
def test_one_iter(self):
w = self.__init_model()
print("before training, coef: {}, intercept: {}".format(
self.coef_, self.intercept_))
self.assertEqual(self.coef_.shape[0], self.guest_X.shape[1])
grad, loss = self.gradient_operator.compute(
self.values,
coef=self.coef_,
intercept=self.intercept_,
fit_intercept=self.fit_intercept)
loss_norm = self.updater.loss_norm(self.coef_)
loss = loss + loss_norm
delta_grad = self.optimizer.apply_gradients(grad)
self.update_model(delta_grad)
print("After training, coef: {}, intercept: {}, loss: {}".format(
self.coef_, self.intercept_, loss))
def test_multi_iter(self):
w = self.__init_model()
loss_hist = [100]
for iter_num in range(self.max_iter):
grad, loss = self.gradient_operator.compute(
self.values,
coef=self.coef_,
intercept=self.intercept_,
fit_intercept=self.fit_intercept)
loss_norm = self.updater.loss_norm(self.coef_)
loss = loss + loss_norm
delta_grad = self.optimizer.apply_gradients(grad)
self.update_model(delta_grad)
self.assertTrue(loss <= loss_hist[-1])
loss_hist.append(loss)
print(loss_hist)
def test_host_iter(self):
w = self.__init_host_model()
print("before training, coef: {}, intercept: {}".format(
self.coef_, self.intercept_))
self.assertEqual(self.host_coef_.shape[0], self.host_X.shape[1])
grad, loss = self.gradient_operator.compute(
self.host_values,
coef=self.host_coef_,
intercept=self.intercept_,
fit_intercept=self.fit_intercept)
loss_norm = self.updater.loss_norm(self.coef_)
# print("***********************************************")
# print(loss, loss_norm)
self.assertTrue(loss is None)
def update_model(self, gradient):
LOGGER.debug(
"In update_model function, shape of coef: {}, shape of gradient: {}"
.format(np.shape(self.coef_), np.shape(gradient)))
if self.fit_intercept:
if self.updater is not None:
self.coef_ = self.updater.update_coef(self.coef_,
gradient[:-1])
else:
self.coef_ = self.coef_ - gradient[:-1]
self.intercept_ -= gradient[-1]
else:
if self.updater is not None:
self.coef_ = self.updater.update_coef(self.coef_, gradient)
else:
self.coef_ = self.coef_ - gradient
class HomoLRHost(BaseLogisticRegression):
def __init__(self, params: LogisticParam):
super(HomoLRHost, self).__init__(params)
self.learning_rate = params.learning_rate
self.batch_size = params.batch_size
self.encrypt_params = params.encrypt_param
if self.encrypt_params.method in [consts.PAILLIER]:
self.use_encrypt = True
else:
self.use_encrypt = False
if self.use_encrypt and params.penalty != consts.L2_PENALTY:
raise RuntimeError("Encrypted h**o-lr supports L2 penalty only")
if self.use_encrypt:
self.gradient_operator = TaylorLogisticGradient()
self.re_encrypt_batches = params.re_encrypt_batches
else:
self.gradient_operator = LogisticGradient()
self.aggregator = HomoFederatedAggregator()
self.party_weight = params.party_weight
self.optimizer = Optimizer(learning_rate=self.learning_rate, opt_method_name=params.optimizer)
self.transfer_variable = HomoLRTransferVariable()
self.initializer = Initializer()
self.mini_batch_obj = None
self.evaluator = Evaluation(classi_type=consts.BINARY)
self.classes_ = [0, 1]
self.has_sychronized_encryption = False
def fit(self, data_instances):
LOGGER.info("parameters: alpha: {}, eps: {}, max_iter: {}"
"batch_size: {}".format(self.alpha,
self.eps, self.max_iter, self.batch_size))
self.__init_parameters(data_instances)
w = self.__init_model(data_instances)
for iter_num in range(self.max_iter):
# mini-batch
LOGGER.debug("In iter: {}".format(iter_num))
batch_data_generator = self.mini_batch_obj.mini_batch_data_generator()
batch_num = 0
total_loss = 0
for batch_data in batch_data_generator:
f = functools.partial(self.gradient_operator.compute,
coef=self.coef_,
intercept=self.intercept_,
fit_intercept=self.fit_intercept)
grad_loss = batch_data.mapPartitions(f)
n = grad_loss.count()
if not self.use_encrypt:
grad, loss = grad_loss.reduce(self.aggregator.aggregate_grad_loss)
grad = np.array(grad)
grad /= n
loss /= n
# gradient_regular, loss_regular = self.updater.compute(w)
if self.updater is not None:
loss_norm = self.updater.loss_norm(self.coef_)
total_loss += loss + loss_norm
# LOGGER.debug("iter: {}, grad: {}, loss: {}".format(iter_num, grad, loss))
else:
grad, _ = grad_loss.reduce(self.aggregator.aggregate_grad)
grad = np.array(grad)
grad /= n
# gradient_regular = self.updater.gradient_norm(w)
# grad += gradient_regular
# grad = np.array(grad)
self.update_model(grad)
w = self.merge_model()
batch_num += 1
if self.use_encrypt and batch_num % self.re_encrypt_batches == 0:
to_encrypt_model_id = self.transfer_variable.generate_transferid(
self.transfer_variable.to_encrypt_model, iter_num, batch_num
)
federation.remote(w,
name=self.transfer_variable.to_encrypt_model.name,
tag=to_encrypt_model_id,
role=consts.ARBITER,
idx=0)
re_encrypted_model_id = self.transfer_variable.generate_transferid(
self.transfer_variable.re_encrypted_model, iter_num, batch_num
)
LOGGER.debug("re_encrypted_model_id: {}".format(re_encrypted_model_id))
w = federation.get(name=self.transfer_variable.re_encrypted_model.name,
tag=re_encrypted_model_id,
idx=0)
w = np.array(w)
self.set_coef_(w)
model_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.host_model, iter_num)
federation.remote(w,
name=self.transfer_variable.host_model.name,
tag=model_transfer_id,
role=consts.ARBITER,
idx=0)
if not self.use_encrypt:
loss_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.host_loss, iter_num)
federation.remote(total_loss,
name=self.transfer_variable.host_loss.name,
tag=loss_transfer_id,
role=consts.ARBITER,
idx=0)
LOGGER.debug("model and loss sent")
final_model_id = self.transfer_variable.generate_transferid(
self.transfer_variable.final_model, iter_num)
w = federation.get(name=self.transfer_variable.final_model.name,
tag=final_model_id,
idx=0)
w = np.array(w)
# LOGGER.debug("Recevide model from arbiter, model: {}".format(w))
self.set_coef_(w)
converge_flag_id = self.transfer_variable.generate_transferid(
self.transfer_variable.converge_flag, iter_num)
converge_flag = federation.get(name=self.transfer_variable.converge_flag.name,
tag=converge_flag_id,
idx=0)
self.n_iter_ = iter_num
LOGGER.debug("converge_flag: {}".format(converge_flag))
if converge_flag:
break
# self.save_model()
def __init_parameters(self, data_instances):
party_weight_id = self.transfer_variable.generate_transferid(
self.transfer_variable.host_party_weight
)
# LOGGER.debug("party_weight_id: {}".format(party_weight_id))
federation.remote(self.party_weight,
name=self.transfer_variable.host_party_weight.name,
tag=party_weight_id,
role=consts.ARBITER,
idx=0)
self.__synchronize_encryption()
# Send re-encrypt times
self.mini_batch_obj = MiniBatch(data_inst=data_instances, batch_size=self.batch_size)
if self.use_encrypt:
# LOGGER.debug("Use encryption, send re_encrypt_times")
total_batch_num = self.mini_batch_obj.batch_nums
re_encrypt_times = total_batch_num // self.re_encrypt_batches
transfer_id = self.transfer_variable.generate_transferid(self.transfer_variable.re_encrypt_times)
federation.remote(re_encrypt_times,
name=self.transfer_variable.re_encrypt_times.name,
tag=transfer_id,
role=consts.ARBITER,
idx=0)
LOGGER.info("sent re_encrypt_times: {}".format(re_encrypt_times))
def __synchronize_encryption(self):
"""
Communicate with hosts. Specify whether use encryption or not and transfer the public keys.
"""
# Send if this host use encryption or not
use_encryption_id = self.transfer_variable.generate_transferid(
self.transfer_variable.use_encrypt
)
federation.remote(self.use_encrypt,
name=self.transfer_variable.use_encrypt.name,
tag=use_encryption_id,
role=consts.ARBITER,
idx=0)
# Set public key
if self.use_encrypt:
pubkey_id = self.transfer_variable.generate_transferid(self.transfer_variable.paillier_pubkey)
pubkey = federation.get(name=self.transfer_variable.paillier_pubkey.name,
tag=pubkey_id,
idx=0)
self.encrypt_operator.set_public_key(pubkey)
LOGGER.info("Finish synchronized ecryption")
self.has_sychronized_encryption = True
def predict(self, data_instances, predict_param):
if not self.has_sychronized_encryption:
self.__synchronize_encryption()
self.__load_arbiter_model()
else:
LOGGER.info("in predict, has synchronize encryption information")
wx = self.compute_wx(data_instances, self.coef_, self.intercept_)
if self.use_encrypt:
encrypted_wx_id = self.transfer_variable.generate_transferid(self.transfer_variable.predict_wx)
# LOGGER.debug("predict_wd_id: {}".format(encrypted_wx_id))
federation.remote(wx,
name=self.transfer_variable.predict_wx.name,
tag=encrypted_wx_id,
role=consts.ARBITER,
idx=0)
predict_result_id = self.transfer_variable.generate_transferid(self.transfer_variable.predict_result)
# LOGGER.debug("predict_result_id: {}".format(predict_result_id))
predict_result = federation.get(name=self.transfer_variable.predict_result.name,
tag=predict_result_id,
idx=0)
# local_predict_table = predict_result.collect()
predict_result_table = predict_result.join(data_instances, lambda p, d: (d.label, None, p))
else:
pred_prob = wx.mapValues(lambda x: activation.sigmoid(x))
pred_label = self.classified(pred_prob, predict_param.threshold)
if predict_param.with_proba:
predict_result = data_instances.mapValues(lambda x: x.label)
predict_result = predict_result.join(pred_prob, lambda x, y: (x, y))
else:
predict_result = data_instances.mapValues(lambda x: (x.label, None))
predict_result_table = predict_result.join(pred_label, lambda x, y: (x[0], x[1], y))
return predict_result_table
def __init_model(self, data_instances):
model_shape = self.get_features_shape(data_instances)
w = self.initializer.init_model(model_shape, init_params=self.init_param_obj)
w = self.encrypt_operator.encrypt_list(w)
w = np.array(w)
# LOGGER.debug("self use encryption: {}, w: {}, type of w: {}".format(self.use_encrypt, w, type(w)))
if self.fit_intercept:
self.coef_ = w[:-1]
self.intercept_ = w[-1]
else:
self.coef_ = w
self.intercept_ = 0
# LOGGER.debug("Type of coef: {}".format(type(self.coef_)))
return w
def __load_arbiter_model(self):
final_model_id = self.transfer_variable.generate_transferid(self.transfer_variable.final_model, "predict")
final_model = federation.get(name=self.transfer_variable.final_model.name,
tag=final_model_id,
idx=0)
self.set_coef_(final_model)
class HomoLRGuest(BaseLogisticRegression):
def __init__(self, params: LogisticParam):
super(HomoLRGuest, self).__init__(params)
self.learning_rate = params.learning_rate
self.aggregator = HomoFederatedAggregator
self.gradient_operator = LogisticGradient()
self.party_weight = params.party_weight
self.optimizer = Optimizer(learning_rate=self.learning_rate,
opt_method_name=params.optimizer)
self.transfer_variable = HomoLRTransferVariable()
self.initializer = Initializer()
self.classes_ = [0, 1]
self.evaluator = Evaluation()
self.header = []
self.penalty = params.penalty
self.loss_history = []
self.is_converged = False
def fit(self, data_instances):
self._abnormal_detection(data_instances)
self.header = data_instances.schema.get(
'header') # ['x1', 'x2', 'x3' ... ]
self.__init_parameters()
self.__init_model(data_instances)
mini_batch_obj = MiniBatch(data_inst=data_instances,
batch_size=self.batch_size)
for iter_num in range(self.max_iter):
# mini-batch
batch_data_generator = mini_batch_obj.mini_batch_data_generator()
total_loss = 0
batch_num = 0
for batch_data in batch_data_generator:
n = batch_data.count()
f = functools.partial(self.gradient_operator.compute,
coef=self.coef_,
intercept=self.intercept_,
fit_intercept=self.fit_intercept)
grad_loss = batch_data.mapPartitions(f)
grad, loss = grad_loss.reduce(
self.aggregator.aggregate_grad_loss)
grad /= n
loss /= n
if self.updater is not None:
loss_norm = self.updater.loss_norm(self.coef_)
total_loss += (loss + loss_norm)
delta_grad = self.optimizer.apply_gradients(grad)
self.update_model(delta_grad)
batch_num += 1
total_loss /= batch_num
w = self.merge_model()
self.loss_history.append(total_loss)
LOGGER.info("iter: {}, loss: {}".format(iter_num, total_loss))
# send model
model_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_model, iter_num)
federation.remote(w,
name=self.transfer_variable.guest_model.name,
tag=model_transfer_id,
role=consts.ARBITER,
idx=0)
# send loss
loss_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_loss, iter_num)
federation.remote(total_loss,
name=self.transfer_variable.guest_loss.name,
tag=loss_transfer_id,
role=consts.ARBITER,
idx=0)
# recv model
model_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.final_model, iter_num)
w = federation.get(name=self.transfer_variable.final_model.name,
tag=model_transfer_id,
idx=0)
w = np.array(w)
self.set_coef_(w)
# recv converge flag
converge_flag_id = self.transfer_variable.generate_transferid(
self.transfer_variable.converge_flag, iter_num)
converge_flag = federation.get(
name=self.transfer_variable.converge_flag.name,
tag=converge_flag_id,
idx=0)
self.n_iter_ = iter_num
LOGGER.debug("converge flag is :{}".format(converge_flag))
if converge_flag:
self.is_converged = True
break
self.show_meta()
self.show_model()
LOGGER.debug("in fit self coef: {}".format(self.coef_))
return data_instances
def __init_parameters(self):
party_weight_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_party_weight)
federation.remote(self.party_weight,
name=self.transfer_variable.guest_party_weight.name,
tag=party_weight_id,
role=consts.ARBITER,
idx=0)
# LOGGER.debug("party weight sent")
LOGGER.info("Finish initialize parameters")
def __init_model(self, data_instances):
model_shape = data_overview.get_features_shape(data_instances)
LOGGER.info("Initialized model shape is {}".format(model_shape))
w = self.initializer.init_model(model_shape,
init_params=self.init_param_obj)
if self.fit_intercept:
self.coef_ = w[:-1]
self.intercept_ = w[-1]
else:
self.coef_ = w
self.intercept_ = 0
# LOGGER.debug("Initialed model")
return w
def predict(self, data_instances, predict_param):
LOGGER.debug("coef: {}, intercept: {}".format(self.coef_,
self.intercept_))
wx = self.compute_wx(data_instances, self.coef_, self.intercept_)
pred_prob = wx.mapValues(lambda x: activation.sigmoid(x))
pred_label = self.classified(pred_prob, predict_param.threshold)
if predict_param.with_proba:
predict_result = data_instances.mapValues(lambda x: x.label)
predict_result = predict_result.join(pred_prob, lambda x, y:
(x, y))
else:
predict_result = data_instances.mapValues(lambda x:
(x.label, None))
predict_result = predict_result.join(pred_label, lambda x, y:
(x[0], x[1], y))
return predict_result
def set_flowid(self, flowid=0):
self.transfer_variable.set_flowid(flowid)
class HomoLRGuest(HomoLRBase):
def __init__(self):
super(HomoLRGuest, self).__init__()
self.aggregator = HomoFederatedAggregator
self.gradient_operator = LogisticGradient()
self.initializer = Initializer()
self.classes_ = [0, 1]
self.evaluator = Evaluation()
self.loss_history = []
self.is_converged = False
self.role = consts.GUEST
def fit(self, data_instances):
if not self.need_run:
return data_instances
self._abnormal_detection(data_instances)
self.init_schema(data_instances)
self.__init_parameters()
self.__init_model(data_instances)
mini_batch_obj = MiniBatch(data_inst=data_instances,
batch_size=self.batch_size)
for iter_num in range(self.max_iter):
# mini-batch
batch_data_generator = mini_batch_obj.mini_batch_data_generator()
total_loss = 0
batch_num = 0
for batch_data in batch_data_generator:
n = batch_data.count()
f = functools.partial(self.gradient_operator.compute,
coef=self.coef_,
intercept=self.intercept_,
fit_intercept=self.fit_intercept)
grad_loss = batch_data.mapPartitions(f)
grad, loss = grad_loss.reduce(
self.aggregator.aggregate_grad_loss)
grad /= n
loss /= n
if self.updater is not None:
loss_norm = self.updater.loss_norm(self.coef_)
total_loss += (loss + loss_norm)
delta_grad = self.optimizer.apply_gradients(grad)
self.update_model(delta_grad)
batch_num += 1
total_loss /= batch_num
# if not self.use_loss:
# total_loss = np.linalg.norm(self.coef_)
w = self.merge_model()
if not self.need_one_vs_rest:
metric_meta = MetricMeta(name='train',
metric_type="LOSS",
extra_metas={
"unit_name": "iters",
})
# metric_name = self.get_metric_name('loss')
self.callback_meta(metric_name='loss',
metric_namespace='train',
metric_meta=metric_meta)
self.callback_metric(
metric_name='loss',
metric_namespace='train',
metric_data=[Metric(iter_num, total_loss)])
self.loss_history.append(total_loss)
LOGGER.info("iter: {}, loss: {}".format(iter_num, total_loss))
# send model
model_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_model, iter_num)
LOGGER.debug("Start to remote model: {}, transfer_id: {}".format(
w, model_transfer_id))
federation.remote(w,
name=self.transfer_variable.guest_model.name,
tag=model_transfer_id,
role=consts.ARBITER,
idx=0)
# send loss
# if self.use_loss:
loss_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_loss, iter_num)
LOGGER.debug(
"Start to remote total_loss: {}, transfer_id: {}".format(
total_loss, loss_transfer_id))
federation.remote(total_loss,
name=self.transfer_variable.guest_loss.name,
tag=loss_transfer_id,
role=consts.ARBITER,
idx=0)
# recv model
model_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.final_model, iter_num)
w = federation.get(name=self.transfer_variable.final_model.name,
tag=model_transfer_id,
idx=0)
w = np.array(w)
self.set_coef_(w)
# recv converge flag
converge_flag_id = self.transfer_variable.generate_transferid(
self.transfer_variable.converge_flag, iter_num)
converge_flag = federation.get(
name=self.transfer_variable.converge_flag.name,
tag=converge_flag_id,
idx=0)
self.n_iter_ = iter_num
LOGGER.debug("converge flag is :{}".format(converge_flag))
if converge_flag:
self.is_converged = True
break
def __init_parameters(self):
party_weight_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_party_weight)
LOGGER.debug(
"Start to remote party_weight: {}, transfer_id: {}".format(
self.party_weight, party_weight_id))
federation.remote(self.party_weight,
name=self.transfer_variable.guest_party_weight.name,
tag=party_weight_id,
role=consts.ARBITER,
idx=0)
# LOGGER.debug("party weight sent")
LOGGER.info("Finish initialize parameters")
def __init_model(self, data_instances):
model_shape = data_overview.get_features_shape(data_instances)
LOGGER.info("Initialized model shape is {}".format(model_shape))
w = self.initializer.init_model(model_shape,
init_params=self.init_param_obj)
if self.fit_intercept:
self.coef_ = w[:-1]
self.intercept_ = w[-1]
else:
self.coef_ = w
self.intercept_ = 0
# LOGGER.debug("Initialed model")
return w
def predict(self, data_instances):
if not self.need_run:
return data_instances
LOGGER.debug(
"homo_lr guest need run predict, coef: {}, instercept: {}".format(
len(self.coef_), self.intercept_))
wx = self.compute_wx(data_instances, self.coef_, self.intercept_)
pred_prob = wx.mapValues(lambda x: activation.sigmoid(x))
pred_label = self.classified(pred_prob, self.predict_param.threshold)
predict_result = data_instances.mapValues(lambda x: x.label)
predict_result = predict_result.join(pred_prob, lambda x, y: (x, y))
predict_result = predict_result.join(
pred_label,
lambda x, y: [x[0], y, x[1], {
"1": x[1],
"0": (1 - x[1])
}])
return predict_result
class HomoLRGuest(BaseLogisticRegression):
def __init__(self, params: LogisticParam):
"""
:param penalty: l1 or l2
:param alpha:
:param lr:
:param eps:
:param max_iter:
:param optim_method: must be in ['sgd', 'RMSProp' ,'Adam', 'AdaGrad']
:param batch_size: only work when otpim_method is mini-batch, represent for mini-batch's size
"""
super(HomoLRGuest, self).__init__(params)
self.learning_rate = params.learning_rate
self.aggregator = HomoFederatedAggregator
self.gradient_operator = LogisticGradient()
self.party_weight = params.party_weight
self.optimizer = Optimizer(learning_rate=self.learning_rate,
opt_method_name=params.optimizer)
self.transfer_variable = HomoLRTransferVariable()
self.initializer = Initializer()
self.classes_ = [0, 1]
self.evaluator = Evaluation()
def fit(self, data_instances):
LOGGER.info("parameters: alpha: {}, eps: {}, max_iter: {}"
"batch_size: {}".format(self.alpha, self.eps,
self.max_iter, self.batch_size))
self.__init_parameters()
w = self.__init_model(data_instances)
mini_batch_obj = MiniBatch(data_inst=data_instances,
batch_size=self.batch_size)
for iter_num in range(self.max_iter):
# mini-batch
# LOGGER.debug("Enter iter_num: {}".format(iter_num))
batch_data_generator = mini_batch_obj.mini_batch_data_generator()
total_loss = 0
batch_num = 0
for batch_data in batch_data_generator:
f = functools.partial(self.gradient_operator.compute,
coef=self.coef_,
intercept=self.intercept_,
fit_intercept=self.fit_intercept)
grad_loss = batch_data.mapPartitions(f)
n = grad_loss.count()
grad, loss = grad_loss.reduce(
self.aggregator.aggregate_grad_loss)
grad /= n
loss /= n
if self.updater is not None:
loss_norm = self.updater.loss_norm(self.coef_)
total_loss += (loss + loss_norm)
# LOGGER.debug("before update: {}".format(grad))
delta_grad = self.optimizer.apply_gradients(grad)
# LOGGER.debug("after apply: {}".format(delta_grad))
self.update_model(delta_grad)
batch_num += 1
total_loss /= batch_num
w = self.merge_model()
LOGGER.info("iter: {}, loss: {}".format(iter_num, total_loss))
# send model
model_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_model, iter_num)
federation.remote(w,
name=self.transfer_variable.guest_model.name,
tag=model_transfer_id,
role=consts.ARBITER,
idx=0)
# send loss
loss_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_loss, iter_num)
federation.remote(total_loss,
name=self.transfer_variable.guest_loss.name,
tag=loss_transfer_id,
role=consts.ARBITER,
idx=0)
# recv model
model_transfer_id = self.transfer_variable.generate_transferid(
self.transfer_variable.final_model, iter_num)
w = federation.get(name=self.transfer_variable.final_model.name,
tag=model_transfer_id,
idx=0)
w = np.array(w)
# LOGGER.debug("Received final model: {}".format(w))
self.set_coef_(w)
# recv converge flag
converge_flag_id = self.transfer_variable.generate_transferid(
self.transfer_variable.converge_flag, iter_num)
converge_flag = federation.get(
name=self.transfer_variable.converge_flag.name,
tag=converge_flag_id,
idx=0)
self.n_iter_ = iter_num
LOGGER.debug("converge flag is :{}".format(converge_flag))
if converge_flag:
# self.save_model(w)
break
# LOGGER.info("trainning finish, final coef: {}, final intercept: {}".format(
# self.coef_, self.intercept_))
def __init_parameters(self):
party_weight_id = self.transfer_variable.generate_transferid(
self.transfer_variable.guest_party_weight)
federation.remote(self.party_weight,
name=self.transfer_variable.guest_party_weight.name,
tag=party_weight_id,
role=consts.ARBITER,
idx=0)
# LOGGER.debug("party weight sent")
LOGGER.info("Finish initialize parameters")
def __init_model(self, data_instances):
model_shape = self.get_features_shape(data_instances)
LOGGER.info("Initialized model shape is {}".format(model_shape))
w = self.initializer.init_model(model_shape,
init_params=self.init_param_obj)
if self.fit_intercept:
self.coef_ = w[:-1]
self.intercept_ = w[-1]
else:
self.coef_ = w
self.intercept_ = 0
# LOGGER.debug("Initialed model")
return w
def predict(self, data_instances, predict_param):
wx = self.compute_wx(data_instances, self.coef_, self.intercept_)
pred_prob = wx.mapValues(lambda x: activation.sigmoid(x))
pred_label = self.classified(pred_prob, predict_param.threshold)
if predict_param.with_proba:
predict_result = data_instances.mapValues(lambda x: x.label)
predict_result = predict_result.join(pred_prob, lambda x, y:
(x, y))
else:
predict_result = data_instances.mapValues(lambda x:
(x.label, None))
predict_result = predict_result.join(pred_label, lambda x, y:
(x[0], x[1], y))
return predict_result
def set_flowid(self, flowid=0):
self.transfer_variable.set_flowid(flowid)