def get_loss(loss):
"""获取使用的损失函数实例.
Arguments:
loss: str or classicML.losses.Loss 实例,
损失函数.
"""
if isinstance(loss, str):
if loss in ('mse', 'mean_squared_error'):
return losses.MeanSquaredError()
elif loss == 'log_likelihood':
return losses.LogLikelihood()
elif loss == 'binary_crossentropy':
return losses.BinaryCrossentropy()
elif loss == 'categorical_crossentropy':
return losses.CategoricalCrossentropy()
elif loss == 'crossentropy':
return losses.Crossentropy()
else:
CLASSICML_LOGGER.warn('你没有输入损失函数或者输入的损失函数不正确, 将使用默认的损失函数')
return losses.Crossentropy()
elif isinstance(loss, losses.Loss):
return loss
else:
CLASSICML_LOGGER.warn('你没有输入损失函数或者输入的损失函数不正确, 将使用默认的损失函数')
return losses.Crossentropy()
def get_kernel(kernel, gamma):
"""获取使用的核函数实例.
Arguments:
kernel: str or classicML.kernels.Kernel 实例,
核函数.
gamma: float, 核函数系数.
Raises:
AttributeError: 模型编译的参数输入错误.
"""
if isinstance(kernel, str):
if kernel == 'linear':
return kernels.Linear()
elif kernel == 'rbf':
return kernels.RBF(gamma=gamma)
elif kernel == 'gaussian':
return kernels.Gaussian(gamma=gamma)
elif kernel == 'poly':
return kernels.Polynomial(gamma=gamma)
elif kernel == 'sigmoid':
return kernels.Sigmoid(gamma=gamma)
else:
CLASSICML_LOGGER.error('核函数调用错误')
raise AttributeError
elif isinstance(kernel, kernels.Kernel):
return kernel
else:
CLASSICML_LOGGER.error('核函数调用错误')
raise AttributeError
def plot_support_vector_classifier(svc, x, y, x_label=None, y_label=None):
"""可视化支持向量分类器二维示意图.
Arguments:
svc: classicML.models.SVC, 支持向量分类器实例.
x: numpy.ndarray, array-like, 特征数据.
y: numpy.ndarray, array-like, 标签.
x_label: str, default=None,
横轴的标签.
y_label: str, default=None,
纵轴的标签.
Raises:
ValueError: 模型没有训练的错误.
"""
if svc.is_trained is False and svc.is_loaded is False:
CLASSICML_LOGGER.error('模型没有训练')
raise ValueError('你必须先进行训练')
ax = plt.subplot()
_set_svc_axis(ax)
# 绘制样本点和支持向量
_plot_sample_and_support_scatter(x, y, svc.support)
# 绘制决策边界
CS = _plot_decision_boundary(svc)
_svc_plot_config(CS, svc.kernel.name, svc.C, x_label, y_label)
plt.show()
def get_initializer(initializer, seed):
"""获取使用的初始化器实例.
Arguments:
initializer: str or classicML.initializers.Initializer 实例,
初始化器.
seed: int, 初始化器的随机种子.
Raises:
AttributeError: 模型编译的参数输入错误.
"""
if isinstance(initializer, str):
if initializer == 'random_normal':
return initializers.RandomNormal(seed=seed)
elif initializer == 'he_normal':
return initializers.HeNormal(seed=seed)
elif initializer in ('xavier_normal', 'glorot_normal'):
return initializers.XavierNormal(seed=seed)
elif initializer == 'rbf_normal':
return initializers.RBFNormal(seed=seed)
else:
CLASSICML_LOGGER.error('初始化器调用错误')
raise AttributeError
elif isinstance(initializer, initializers.Initializer):
return initializer
elif initializer is None:
return initializers.RandomNormal(seed=seed)
else:
CLASSICML_LOGGER.error('初始化器调用错误')
raise AttributeError
def save_weights(self, filepath):
"""将模型权重保存为一个HDF5文件.
Arguments:
filepath: str, 权重文件保存的路径.
Raises:
TypeError: 模型权重保存失败.
Notes:
模型将不会保存关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath,
mode='w',
model_name='RadialBasisFunctionNetwork')
# 保存模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
compile_ds.attrs['hidden_units'] = self.hidden_units
compile_ds.attrs['optimizer'] = self.optimizer.name
compile_ds.attrs['loss'] = self.loss.name
compile_ds.attrs['metric'] = self.metric.name
for parameter in self.parameters:
weights_ds.attrs[parameter] = self.parameters[parameter]
except TypeError:
CLASSICML_LOGGER.error('模型权重保存失败, 请检查文件是否损坏')
raise TypeError('模型权重保存失败')
def save_weights(self, filepath):
"""将模型权重保存为一个HDF5文件.
Arguments:
filepath: str, 权重文件保存的路径.
Raises:
TypeError: 模型权重保存失败.
Notes:
模型将不会保存关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath,
mode='w',
model_name='AveragedOneDependentEstimator')
# 保存模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
compile_ds.attrs['smoothing'] = self.smoothing
compile_ds.attrs['m'] = self.m
weights_ds.attrs['_attribute_list'] = np.void(
dumps(self._attribute_list))
except TypeError:
CLASSICML_LOGGER.error('模型权重保存失败, 请检查文件是否损坏')
raise TypeError('模型权重保存失败')
def save_weights(self, filepath):
"""将模型权重保存为一个HDF5文件.
Arguments:
filepath: str, 权重文件保存的路径.
Raises:
TypeError: 模型权重保存失败.
Notes:
模型将不会保存关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath, mode='w', model_name='SupportVectorClassifier')
# 保存模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
compile_ds.attrs['C'] = self.C
compile_ds.attrs['kernel'] = self.kernel.name
compile_ds.attrs['gamma'] = self.gamma
compile_ds.attrs['tol'] = self.tol
weights_ds.attrs['support'] = self.support
weights_ds.attrs['support_vector'] = self.support_vector
weights_ds.attrs['support_alpha'] = self.support_alpha
weights_ds.attrs['support_y'] = np.asarray(self.support_y,
dtype=np.float64)
weights_ds.attrs['b'] = np.float64(self.b)
except TypeError:
CLASSICML_LOGGER.error('模型权重保存失败, 请检查文件是否损坏')
raise TypeError('模型权重保存失败')
def plot_bayes(bayes, x, y):
"""可视化朴素贝叶斯分类器或超父独依赖估计器的二维示意图.
Arguments:
bayes: classicML.models.NB or classicML.models.SPODE,
朴素贝叶斯分类器或超父独依赖估计器实例.
x: numpy.ndarray, array-like, 特征数据.
y: numpy.ndarray, array-like, 标签.
Raises:
ValueError: 模型没有训练的错误.
"""
if bayes.is_trained is False and bayes.is_loaded is False:
CLASSICML_LOGGER.error('模型没有训练')
raise ValueError('你必须先进行训练')
ax = plt.subplot()
_set_bayes_axis(ax)
# 绘制背景分界图
_plot_background(bayes, x)
# 绘制样本点
x_label, y_label = _plot_scatter(x, y)
_bayes_plot_config(x_label, y_label)
plt.show()
def predict(self, x):
"""使用决策树分类器进行预测.
Arguments:
x: numpy.ndarray or pandas.DataFrame, array-like,
特征数据.
Returns:
DecisionTreeClassifier预测的结果.
Raises:
ValueError: 模型没有训练的错误.
"""
if self.is_trained is False and self.is_loaded is False:
CLASSICML_LOGGER.error('模型没有训练')
raise ValueError('你必须先进行训练')
# 修正数据类型.
if isinstance(x, list):
x = np.expand_dims(x, axis=0)
elif isinstance(x, pd.DataFrame):
x = x.values
elif isinstance(x, pd.Series):
x = np.expand_dims(x.values, axis=0)
y_pred = list()
for feature in x:
y_pred.append(self._predict(feature, self.tree))
return y_pred
def load_weights(self, filepath):
"""加载模型参数.
Arguments:
filepath: str, 权重文件加载的路径.
Raises:
KeyError: 模型权重加载失败.
Notes:
模型将不会加载关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath, mode='r', model_name='DecisionTreeClassifier')
# 加载模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
self.criterion = compile_ds.attrs['criterion']
if self.pruner is not None:
self.pruner = get_pruner(compile_ds.attrs['pruning'])
self.tree = loads(weights_ds.attrs['tree'].tobytes())
# 标记加载完成
self.is_loaded = True
except KeyError:
CLASSICML_LOGGER.error('模型权重加载失败, 请检查文件是否损坏')
raise KeyError('模型权重加载失败')
def predict(self, x):
"""模型进行预测.
Arguments:
x: numpy.ndarray, array-like, 特征数据.
Returns:
LDA预测的标签张量.
Raises:
ValueError: 模型没有训练的错误.
"""
if self.is_trained is False and self.is_loaded is False:
CLASSICML_LOGGER.error('模型没有权重')
raise ValueError('你必须先进行训练')
coord = np.dot(x, self.w.T)
center_0 = np.dot(self.w, self.mu_0.T)
center_1 = np.dot(self.w, self.mu_1.T)
y_pred = np.abs(coord - center_0) > np.abs(coord - center_1)
y_pred = np.squeeze(y_pred)
y_pred = y_pred.astype(int)
return y_pred
def save_weights(self, filepath):
"""将模型权重保存为一个HDF5文件.
Arguments:
filepath: str, 权重文件保存的路径.
Raises:
TypeError: 模型权重保存失败.
Notes:
模型将不会保存关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath, mode='w', model_name='LogisticRegression')
# 保存模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
compile_ds.attrs['optimizer'] = self.optimizer.name
compile_ds.attrs['loss'] = self.loss.name
compile_ds.attrs['metric'] = self.metric.name
weights_ds.attrs['beta'] = self.beta
except TypeError:
CLASSICML_LOGGER.error('模型权重保存失败, 请检查文件是否损坏')
raise TypeError('模型权重保存失败')
def load_weights(self, filepath):
"""加载模型参数.
Arguments:
filepath: str, 权重文件加载的路径.
Raises:
KeyError: 模型权重加载失败.
Notes:
模型将不会加载关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath, mode='r', model_name='SupportVectorClassifier')
# 加载模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
self.C = compile_ds.attrs['C']
self.gamma = compile_ds.attrs['gamma']
self.kernel = get_kernel(compile_ds.attrs['kernel'], self.gamma)
self.tol = compile_ds.attrs['tol']
self.support = weights_ds.attrs['support']
self.support_vector = weights_ds.attrs['support_vector']
self.support_alpha = weights_ds.attrs['support_alpha']
self.support_y = weights_ds.attrs['support_y']
self.b = weights_ds.attrs['b']
# 标记加载完成
self.is_loaded = True
except KeyError:
CLASSICML_LOGGER.error('模型权重加载失败, 请检查文件是否损坏')
raise KeyError('模型权重加载失败')
def wrapper(*args, **kwargs):
start_time = time.perf_counter() # 注意将记录time.sleep()的时间
return_values = function(*args, **kwargs)
end_time = time.perf_counter()
CLASSICML_LOGGER.info('耗时 {:.5f} s'.format(end_time - start_time))
return return_values
def get_optimizer(optimizer):
"""获取使用的优化器实例.
Arguments:
optimizer: str or classicML.optimizers.Optimizer 实例,
优化器.
Raises:
AttributeError: 模型编译的参数输入错误.
"""
if isinstance(optimizer, str):
if optimizer in ('gd', 'gradient_descent'):
return optimizers.GradientDescent()
elif optimizer in ('newton', 'newton_method'):
return optimizers.NewtonMethod()
elif optimizer in ('sgd', 'stochastic_gradient_descent'):
return optimizers.StochasticGradientDescent()
elif optimizer == 'adam':
return optimizers.Adam()
elif optimizer == 'rbf':
return optimizers.RadialBasisFunctionOptimizer()
elif optimizer == 'SMO':
return optimizers.SequentialMinimalOptimization()
else:
CLASSICML_LOGGER.error('优化器调用错误')
raise AttributeError
elif isinstance(optimizer, optimizers.Optimizer):
return optimizer
else:
CLASSICML_LOGGER.error('优化器调用错误')
raise AttributeError
def predict(self, x, **kwargs):
"""使用平均独依赖估计器进行预测.
Arguments:
x: numpy.ndarray or pandas.DataFrame, array-like, 特征数据.
Returns:
AverageOneDependentEstimator的预测结果.
Raises:
ValueError: 模型没有训练的错误.
"""
if self.is_trained is False and self.is_loaded is False:
CLASSICML_LOGGER.error('模型没有训练')
raise ValueError('你必须先进行训练')
y_pred = list()
if len(x.shape) == 1:
y_pred.append(self._predict(x))
else:
for i in range(x.shape[0]):
x_test = x.iloc[i, :]
y_pred.append(self._predict(x_test))
return y_pred
def load_weights(self, filepath):
"""加载模型参数.
Arguments:
filepath: str, 权重文件加载的路径.
Raises:
KeyError: 模型权重加载失败.
Notes:
模型将不会加载关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath,
mode='r',
model_name='AveragedOneDependentEstimator')
# 加载模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
self.smoothing = compile_ds.attrs['smoothing']
self.m = compile_ds.attrs['m']
self._attribute_list = loads(
weights_ds.attrs['_attribute_list'].tobytes())
# 标记加载完成
self.is_loaded = True
except KeyError:
CLASSICML_LOGGER.error('模型权重加载失败, 请检查文件是否损坏')
raise KeyError('模型权重加载失败')
def load_weights(self, filepath):
"""加载模型参数.
Arguments:
filepath: str, 权重文件加载的路径.
Raises:
KeyError: 模型权重加载失败.
Notes:
模型将不会加载关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath,
mode='r',
model_name='RadialBasisFunctionNetwork')
# 加载模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
self.hidden_units = compile_ds.attrs['hidden_units']
self.optimizer = get_optimizer(compile_ds.attrs['optimizer'])
self.loss = get_loss(compile_ds.attrs['loss'])
self.metric = get_metric(compile_ds.attrs['metric'])
for attr in weights_ds.attrs:
self.parameters[attr] = weights_ds.attrs[attr]
# 标记加载完成
self.is_loaded = True
except KeyError:
CLASSICML_LOGGER.error('模型权重加载失败, 请检查文件是否损坏')
raise KeyError('模型权重加载失败')
def save_weights(self, filepath):
"""将模型权重保存为一个HDF5文件.
Arguments:
filepath: str, 权重文件保存的路径.
Raises:
TypeError: 模型权重保存失败.
Notes:
模型将不会保存关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath, mode='w', model_name='NaiveBayesClassifier')
# 保存模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
compile_ds.attrs['smoothing'] = self.smoothing
weights_ds.attrs['p_0'] = self.p_0
weights_ds.attrs['p_1'] = self.p_1
weights_ds.attrs['pi_0'] = np.void(dumps(self.pi_0))
weights_ds.attrs['pi_1'] = np.void(dumps(self.pi_1))
except TypeError:
CLASSICML_LOGGER.error('模型权重保存失败, 请检查文件是否损坏')
raise TypeError('模型权重保存失败')
def load_weights(self, filepath):
"""加载模型参数.
Arguments:
filepath: str, 权重文件加载的路径.
Raises:
KeyError: 模型权重加载失败.
Notes:
模型将不会加载关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath, mode='r', model_name='NaiveBayesClassifier')
# 加载模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
self.smoothing = compile_ds.attrs['smoothing']
self.p_0 = weights_ds.attrs['p_0']
self.p_1 = weights_ds.attrs['p_1']
self.pi_0 = loads(weights_ds.attrs['pi_0'].tobytes())
self.pi_1 = loads(weights_ds.attrs['pi_1'].tobytes())
# 标记加载完成
self.is_loaded = True
except KeyError:
CLASSICML_LOGGER.error('模型权重加载失败, 请检查文件是否损坏')
raise KeyError('模型权重加载失败')
def predict(self, x, probability=False):
"""使用朴素贝叶斯分类器进行预测.
Arguments:
x: numpy.ndarray or pandas.DataFrame, array-like, 特征数据.
probability: bool, default=False,
是否使用归一化的概率形式.
Returns:
NaiveBayesClassifier的预测结果,
不使用概率形式将返回0或1的标签数组, 使用将返回反正例概率的数组.
Raises:
ValueError: 模型没有训练的错误.
"""
if self.is_trained is False and self.is_loaded is False:
CLASSICML_LOGGER.error('模型没有训练')
raise ValueError('你必须先进行训练')
# 避免下溢进行对数处理.
p_0 = np.log(self.p_0)
p_1 = np.log(self.p_1)
y_pred = list()
if len(x.shape) == 1:
y_pred.append(self._predict(x, p_0, p_1, probability))
else:
for i in range(x.shape[0]):
x_test = x.iloc[i, :]
y_pred.append(self._predict(x_test, p_0, p_1, probability))
return y_pred
def save_weights(self, filepath):
"""将模型权重保存为一个HDF5文件.
Arguments:
filepath: str, 权重文件保存的路径.
Raises:
TypeError: 模型权重保存失败.
References:
- [如何存储原始的二进制数据](https://docs.h5py.org/en/2.3/strings.html)
Notes:
模型将不会保存关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath, mode='w', model_name='DecisionTreeClassifier')
# 保存模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
compile_ds.attrs['criterion'] = self.criterion
if self.pruner is not None:
compile_ds.attrs['pruning'] = self.pruner.name
weights_ds.attrs['tree'] = np.void(dumps(self.tree))
except TypeError:
CLASSICML_LOGGER.error('模型权重保存失败, 请检查文件是否损坏')
raise TypeError('模型权重保存失败')
def predict(self, x):
"""使用分类器进行预测.
Arguments:
x: numpy.ndarray, array-like, 特征数据.
Returns:
SupportVectorClassifier预测的张量数组.
Raises:
ValueError: 模型没有训练的错误.
"""
if self.is_trained is False and self.is_loaded is False:
CLASSICML_LOGGER.error('模型没有训练')
raise ValueError('你必须先进行训练')
number_of_sample = x.shape[0]
y_pred = np.ones((number_of_sample, ), dtype=int)
for sample in range(number_of_sample):
kappa = self.kernel(x[sample], self.support_vector)
fx = np.matmul(
(self.support_alpha.reshape(-1, 1) * self.support_y).T,
kappa.T) + self.b
if fx < 0:
y_pred[sample] = -1
return y_pred
def load_weights(self, filepath):
"""加载模型参数.
Arguments:
filepath: str, 权重文件加载的路径.
Raises:
KeyError: 模型权重加载失败.
Notes:
模型将不会加载关于优化器的超参数.
"""
# 初始化权重文件.
parameters_gp = io.initialize_weights_file(
filepath=filepath, mode='r', model_name='LogisticRegression')
# 加载模型参数.
try:
compile_ds = parameters_gp['compile']
weights_ds = parameters_gp['weights']
self.optimizer = get_optimizer(compile_ds.attrs['optimizer'])
self.loss = get_loss(compile_ds.attrs['loss'])
self.metric = get_metric(compile_ds.attrs['metric'])
self.beta = weights_ds.attrs['beta']
# 标记加载完成
self.is_loaded = True
except KeyError:
CLASSICML_LOGGER.error('模型权重加载失败, 请检查文件是否损坏')
raise KeyError('模型权重加载失败')
def fit(self, x, y, x_validation=None, y_validation=None):
"""训练决策树分类器.
Arguments:
x: numpy.ndarray or pandas.DataFrame, array-like,
特征数据.
y: numpy.ndarray or pandas.DataFrame, array-like,
标签.
x_validation: numpy.ndarray or pandas.DataFrame, array-like,
剪枝使用的验证特征数据.
y_validation: numpy.ndarray or pandas.DataFrame, array-like,
剪枝使用的验证标签.
Returns:
DecisionTreeClassifier实例.
Raises:
AttributeError: 没有验证集.
"""
if isinstance(x, np.ndarray) and self.attribute_name is None:
CLASSICML_LOGGER.warn(
"属性名称缺失, 请使用pandas.DataFrame; 或检查 self.attributes_name")
if (self.pruner is not None) and (x_validation is None
or y_validation is None):
CLASSICML_LOGGER.error("没有验证集, 无法对决策树进行剪枝")
raise AttributeError('没有验证集')
# 为特征数据添加属性信息.
x = pd.DataFrame(x, columns=self.attribute_name)
x.reset_index(drop=True, inplace=True)
self.generator._x = x
y = pd.Series(y)
y.reset_index(drop=True, inplace=True)
# 为验证数据添加属性信息.
if x_validation is not None:
x_validation = pd.DataFrame(x_validation,
columns=self.attribute_name)
x_validation.reset_index(drop=True, inplace=True)
y_validation = pd.Series(y_validation)
y_validation.reset_index(drop=True, inplace=True)
# 没有使用权重文件, 则生成决策树分类器.
if self.is_loaded is False:
self.tree = self.generator(x, y)
# 进行剪枝.
if self.pruner:
self.tree = self.pruner(x, y, x_validation, y_validation,
self.tree)
# 标记训练完成
self.is_trained = True
return self
def wrapper(*args, **kwargs):
return_values = function(*args, **kwargs)
pid = os.getpid()
current_process = psutil.Process(pid)
process_memory = current_process.memory_full_info()
CLASSICML_LOGGER.info('占用内存 {:.5f} MB'.format(process_memory.uss /
1024 / 1024))
return return_values
def fit(self, x, y):
"""训练平均独依赖估计器.
Arguments:
x: numpy.ndarray or pandas.DataFrame, array-like, 特征数据.
y: numpy.ndarray or pandas.DataFrame, array-like, 标签.
Returns:
AverageOneDependentEstimator实例.
"""
if isinstance(x, np.ndarray) and self.attribute_name is None:
CLASSICML_LOGGER.warn(
"属性名称缺失, 请使用pandas.DataFrame; 或检查 self.attributes_name")
# TODO(Steve R. Sun, tag:code): 暂时没有找到合理的断点续训的理论支持.
self._attribute_list = list()
# 为特征数据添加属性信息.
x = pd.DataFrame(x, columns=self.attribute_name)
x.reset_index(drop=True, inplace=True)
y = pd.Series(y)
y.reset_index(drop=True, inplace=True)
number_of_samples, number_of_attributes = x.shape
# 获取离散属性的全部取值.
discrete_unique_values = dict()
for attribute in range(number_of_attributes):
xi = x.iloc[:, attribute]
if (type_of_target(xi.values) != 'continuous') and (
pd.value_counts(xi).values > self.m).all():
discrete_unique_values.update(
{x.columns[attribute]: xi.unique()})
# 每个属性作为超父类构建SPODE.
for index, key in enumerate(discrete_unique_values.keys()):
self.super_parent_name = key
super(AveragedOneDependentEstimator, self).fit(x, y)
current_attribute_list = self._list_of_p_c
self._attribute_list.append(current_attribute_list)
self.is_trained = True
return self
def _parse(model_name, fp):
"""解析描述信息组, 核验文件.
Arguments:
model_name: str, 模型的名称.
fp: h5py._hl.files.File, 文件指针.
"""
description_gp = fp['description']
file_cml_version = re.findall('\\d+\\.\\d+\\.\\d+', description_gp.attrs['version'])[0]
file_backend_version = 'backend.io' + re.findall('\\d+\\.\\d+', description_gp.attrs['version'])[-1]
if (file_cml_version < min_cml_version) or (file_backend_version < min__version__):
CLASSICML_LOGGER.error('文件核验失败, 模型版本过低')
raise ValueError('文件核验失败')
if description_gp.attrs['model_name'] != model_name:
CLASSICML_LOGGER.error('文件核验失败, 模型不匹配')
raise ValueError('文件核验失败')
def predict(self, x):
"""使用径向基函数网络进行预测.
Arguments:
x: numpy.ndarray, array-like, 特征数据.
Returns:
径向基函数网络预测的概率.
Raises:
ValueError: 模型没有训练的错误.
"""
if self.is_trained is False and self.is_loaded is False:
CLASSICML_LOGGER.error('模型没有训练')
raise ValueError('你必须先进行训练')
y_pred, _ = self.optimizer.forward(x, self.parameters)
return y_pred
def predict(self, x, probability=False):
"""使用超父独依赖估计器进行预测.
Arguments:
x: numpy.ndarray or pandas.DataFrame, array-like, 特征数据.
probability: bool, default=False,
是否使用归一化的概率形式.
Returns:
SuperParentOneDependentEstimator的预测结果,
不使用概率形式将返回0或1的标签数组, 使用将返回反正例概率的数组.
Raises:
ValueError: 模型没有训练的错误.
"""
if self.is_trained is False and self.is_loaded is False:
CLASSICML_LOGGER.error('模型没有训练')
raise ValueError('你必须先进行训练')
y_pred = list()
if len(x.shape) == 1:
p_0, p_1 = self._predict(x)
if probability:
y_pred.append([p_0 / (p_0 + p_1), p_1 / (p_0 + p_1)])
else:
if p_0 > p_1:
y_pred.append(0)
else:
y_pred.append(1)
else:
for i in range(x.shape[0]):
x_test = x.iloc[i, :]
p_0, p_1 = self._predict(x_test)
if probability:
y_pred.append([p_0 / (p_0 + p_1), p_1 / (p_0 + p_1)])
else:
if p_0 > p_1:
y_pred.append(0)
else:
y_pred.append(1)
return y_pred
