一、实验介绍
本篇文章主要研究的是前馈网络的神经网络模型,以及两种前馈神经网络:多层感知器和卷积神经网络。旨在通过感知器、多层感知器和卷积神经网络的实现和对比,帮助大家理解神经网络的基础原理及其应用。
对此我们将开展以下两个实验:带有多层感知器的姓氏分类和使用CNN对姓氏进行分类
二、带有多层感知器的姓氏分类
2.1 多层感知机介绍
多层感知器(MLP)被认为是最基本的神经网络构建模块之一。感知器将数据向量作为输入,计算出一个输出值。在MLP中,许多感知器被分组,以便单个层的输出是一个新的向量,而不是单个输出值。在PyTorch中,正如您稍后将看到的,这只需设置线性层中的输出特性的数量即可完成。MLP的另一个方面是,它将多个层与每个层之间的非线性结合在一起。
最简单的MLP,如图4-2所示,由三个表示阶段和两个线性层组成。第一阶段是输入向量。这是给定给模型的向量。在“示例:对餐馆评论的情绪进行分类”中,输入向量是Yelp评论的一个收缩的one-hot表示。给定输入向量,第一个线性层计算一个隐藏向量——表示的第二阶段。隐藏向量之所以这样被调用,是因为它是位于输入和输出之间的层的输出。我们所说的“层的输出”是什么意思?理解这个的一种方法是隐藏向量中的值是组成该层的不同感知器的输出。使用这个隐藏的向量,第二个线性层计算一个输出向量。在像Yelp评论分类这样的二进制任务中,输出向量仍然可以是1。在多类设置中,将在本实验后面的“示例:带有多层感知器的姓氏分类”一节中看到,输出向量是类数量的大小。虽然在这个例子中,我们只展示了一个隐藏的向量,但是有可能有多个中间阶段,每个阶段产生自己的隐藏向量。最终的隐藏向量总是通过线性层和非线性的组合映射到输出向量。
图4-2 一种具有两个线性层和三个表示阶段(输入向量、隐藏向量和输出向量)的MLP的可视化表示
mlp的力量来自于添加第二个线性层和允许模型学习一个线性分割的的中间表示——该属性的能表示一个直线(或更一般的,一个超平面)可以用来区分数据点落在线(或超平面)的哪一边的。学习具有特定属性的中间表示,如分类任务是线性可分的,这是使用神经网络的最深刻后果之一,也是其建模能力的精髓。
激活函数:在多层感知机中,每个隐藏层和输出层的神经元通常都会使用激活函数来引入非线性。常见的激活函数包括ReLU(Rectified Linear Unit)、Sigmoid、Tanh等,用于增加网络的表达能力,使神经网络可以学习复杂的非线性关系。
反向传播算法:多层感知机通过反向传播算法(Backpropagation)来训练网络。反向传播算法基于梯度下降法,通过计算损失函数对网络参数的偏导数(梯度),然后根据梯度更新网络参数,使得网络的预测结果尽可能接近真实标签。
前向传播:在训练完成后,在推理阶段,输入数据沿着网络的前向传播路径传递,经过隐藏层和输出层,最终得到网络的预测结果。
2.2 MLP实现姓氏分类
2.2.1 数据预处理
姓氏数据集,它收集了来自18个不同国家的10,000个姓氏,这些姓氏是作者从互联网上不同的姓名来源收集的。该数据集将在本课程实验的几个示例中重用,并具有一些使其有趣的属性。第一个性质是它是相当不平衡的。排名前三的课程占数据的60%以上:27%是英语,21%是俄语,14%是阿拉伯语。剩下的15个民族的频率也在下降——这也是语言特有的特性。第二个特点是,在国籍和姓氏正字法(拼写)之间有一种有效和直观的关系。有些拼写变体与原籍国联系非常紧密(比如“O ‘Neill”、“Antonopoulos”、“Nagasawa”或“Zhu”)。
为了创建最终的数据集,我们从一个比课程补充材料中包含的版本处理更少的版本开始,并执行了几个数据集修改操作。第一个目的是减少这种不平衡——原始数据集中70%以上是俄文,这可能是由于抽样偏差或俄文姓氏的增多。为此,我们通过选择标记为俄语的姓氏的随机子集对这个过度代表的类进行子样本。接下来,我们根据国籍对数据集进行分组,并将数据集分为三个部分:70%到训练数据集,15%到验证数据集,最后15%到测试数据集,以便跨这些部分的类标签分布具有可比性。
class SurnameDataset(Dataset):
def __init__(self, surname_df, vectorizer):
"""
参数:
surname_df (pandas.DataFrame): 数据集
vectorizer (SurnameVectorizer): 从数据集实例化的向量化器
"""
self.surname_df = surname_df
self._vectorizer = vectorizer
# 拆分数据集为训练集、验证集和测试集
self.train_df = self.surname_df[self.surname_df.split=='train']
self.train_size = len(self.train_df)
self.val_df = self.surname_df[self.surname_df.split=='val']
self.validation_size = len(self.val_df)
self.test_df = self.surname_df[self.surname_df.split=='test']
self.test_size = len(self.test_df)
self._lookup_dict = {'train': (self.train_df, self.train_size),
'val': (self.val_df, self.validation_size),
'test': (self.test_df, self.test_size)}
self.set_split('train')
# 类别权重
class_counts = surname_df.nationality.value_counts().to_dict()
def sort_key(item):
return self._vectorizer.nationality_vocab.lookup_token(item[0])
sorted_counts = sorted(class_counts.items(), key=sort_key)
frequencies = [count for _, count in sorted_counts]
self.class_weights = 1.0 / torch.tensor(frequencies, dtype=torch.float32)
@classmethod
def load_dataset_and_make_vectorizer(cls, surname_csv):
"""加载数据集并从头创建一个新的向量化器
参数:
surname_csv (str): 数据集的位置
返回:
SurnameDataset的一个实例
"""
surname_df = pd.read_csv(surname_csv)
train_surname_df = surname_df[surname_df.split=='train']
return cls(surname_df, SurnameVectorizer.from_dataframe(train_surname_df))
@classmethod
def load_dataset_and_load_vectorizer(cls, surname_csv, vectorizer_filepath):
"""加载数据集和相应的向量化器。用于向量化器已被缓存以供重用的情况
参数:
surname_csv (str): 数据集的位置
vectorizer_filepath (str): 已保存的向量化器的位置
返回:
SurnameDataset的一个实例
"""
surname_df = pd.read_csv(surname_csv)
vectorizer = cls.load_vectorizer_only(vectorizer_filepath)
return cls(surname_df, vectorizer)
@staticmethod
def load_vectorizer_only(vectorizer_filepath):
"""从文件加载向量化器的静态方法
参数:
vectorizer_filepath (str): 序列化向量化器的位置
返回:
SurnameVectorizer的一个实例
"""
with open(vectorizer_filepath) as fp:
return SurnameVectorizer.from_serializable(json.load(fp))
def save_vectorizer(self, vectorizer_filepath):
"""使用json将向量化器保存到磁盘
参数:
vectorizer_filepath (str): 保存向量化器的位置
"""
with open(vectorizer_filepath, "w") as fp:
json.dump(self._vectorizer.to_serializable(), fp)
def get_vectorizer(self):
""" 返回向量化器 """
return self._vectorizer
def set_split(self, split="train"):
""" 使用数据框中的列选择数据集的拆分 """
self._target_split = split
self._target_df, self._target_size = self._lookup_dict[split]
def __len__(self):
return self._target_size
def __getitem__(self, index):
"""PyTorch数据集的主要入口点方法
参数:
index (int): 数据点的索引
返回:
一个包含数据点的字典:
特征 (x_surname)
标签 (y_nationality)
"""
row = self._target_df.iloc[index]
surname_vector = self._vectorizer.vectorize(row.surname)
nationality_index = self._vectorizer.nationality_vocab.lookup_token(row.nationality)
return {'x_surname': surname_vector,
'y_nationality': nationality_index}
def get_num_batches(self, batch_size):
"""给定批量大小,返回数据集中的批次数量
参数:
batch_size (int)
返回:
数据集中的批次数量
"""
return len(self) // batch_size
def generate_batches(dataset, batch_size, shuffle=True,
drop_last=True, device="cpu"):
"""
一个包装PyTorch DataLoader的生成器函数。它将确保每个张量在正确的设备位置上。
"""
dataloader = DataLoader(dataset=dataset, batch_size=batch_size,
shuffle=shuffle, drop_last=drop_last)
for data_dict in dataloader:
out_data_dict = {}
for name, tensor in data_dict.items():
out_data_dict[name] = data_dict[name].to(device)
yield out_data_dict
2.2.2 词汇表、向量化器和DataLoader
为了使用字符对姓氏进行分类,我们使用词汇表、向量化器和DataLoader将姓氏字符串转换为向量化的minibatches。
class Vocabulary(object):
"""Class to process text and extract vocabulary for mapping"""
def __init__(self, token_to_idx=None, add_unk=True, unk_token="<UNK>"):
"""
Args:
token_to_idx (dict): a pre-existing map of tokens to indices
add_unk (bool): a flag that indicates whether to add the UNK token
unk_token (str): the UNK token to add into the Vocabulary
"""
if token_to_idx is None:
token_to_idx = {}
self._token_to_idx = token_to_idx
self._idx_to_token = {idx: token
for token, idx in self._token_to_idx.items()}
self._add_unk = add_unk
self._unk_token = unk_token
self.unk_index = -1
if add_unk:
self.unk_index = self.add_token(unk_token)
def to_serializable(self):
""" returns a dictionary that can be serialized """
return {'token_to_idx': self._token_to_idx,
'add_unk': self._add_unk,
'unk_token': self._unk_token}
@classmethod
def from_serializable(cls, contents):
""" instantiates the Vocabulary from a serialized dictionary """
return cls(**contents)
def add_token(self, token):
"""Update mapping dicts based on the token.
Args:
token (str): the item to add into the Vocabulary
Returns:
index (int): the integer corresponding to the token
"""
try:
index = self._token_to_idx[token]
except KeyError:
index = len(self._token_to_idx)
self._token_to_idx[token] = index
self._idx_to_token[index] = token
return index
def add_many(self, tokens):
"""Add a list of tokens into the Vocabulary
Args:
tokens (list): a list of string tokens
Returns:
indices (list): a list of indices corresponding to the tokens
"""
return [self.add_token(token) for token in tokens]
def lookup_token(self, token):
"""Retrieve the index associated with the token
or the UNK index if token isn't present.
Args:
token (str): the token to look up
Returns:
index (int): the index corresponding to the token
Notes:
`unk_index` needs to be >=0 (having been added into the Vocabulary)
for the UNK functionality
"""
if self.unk_index >= 0:
return self._token_to_idx.get(token, self.unk_index)
else:
return self._token_to_idx[token]
def lookup_index(self, index):
"""Return the token associated with the index
Args:
index (int): the index to look up
Returns:
token (str): the token corresponding to the index
Raises:
KeyError: if the index is not in the Vocabulary
"""
if index not in self._idx_to_token:
raise KeyError("the index (%d) is not in the Vocabulary" % index)
return self._idx_to_token[index]
def __str__(self):
return "<Vocabulary(size=%d)>" % len(self)
def __len__(self):
return len(self._token_to_idx)
2.2.3 分类模型
class SurnameClassifier(nn.Module):
""" A 2-layer Multilayer Perceptron for classifying surnames """
def __init__(self, input_dim, hidden_dim, output_dim):
"""
Args:
input_dim (int): the size of the input vectors
hidden_dim (int): the output size of the first Linear layer
output_dim (int): the output size of the second Linear layer
"""
super(SurnameClassifier, self).__init__()
self.fc1 = nn.Linear(input_dim, hidden_dim)
self.fc2 = nn.Linear(hidden_dim, output_dim)
def forward(self, x_in, apply_softmax=False):
"""The forward pass of the classifier
Args:
x_in (torch.Tensor): an input data tensor.
x_in.shape should be (batch, input_dim)
apply_softmax (bool): a flag for the softmax activation
should be false if used with the Cross Entropy losses
Returns:
the resulting tensor. tensor.shape should be (batch, output_dim)
"""
intermediate_vector = F.relu(self.fc1(x_in))
prediction_vector = self.fc2(intermediate_vector)
if apply_softmax:
prediction_vector = F.softmax(prediction_vector, dim=1)
return prediction_vector
2.2.4 模型训练
# 设置参数
args = Namespace(
# 数据和路径信息
surname_csv="data/surnames/surnames_with_splits.csv",
vectorizer_file="vectorizer.json",
model_state_file="model.pth",
save_dir="model_storage/ch4/surname_mlp",
# 模型超参数
hidden_dim=300,
# 训练超参数
seed=1337,
num_epochs=100,
early_stopping_criteria=5,
learning_rate=0.001,
batch_size=64,
# 运行时选项
cuda=False,
reload_from_files=False,
expand_filepaths_to_save_dir=True,
)
# 扩展文件路径
if args.expand_filepaths_to_save_dir:
args.vectorizer_file = os.path.join(args.save_dir, args.vectorizer_file)
args.model_state_file = os.path.join(args.save_dir, args.model_state_file)
print("扩展的文件路径: ")
print("\t{}".format(args.vectorizer_file))
print("\t{}".format(args.model_state_file))
# 检查 CUDA 可用性
if not torch.cuda.is_available():
args.cuda = False
args.device = torch.device("cuda" if args.cuda else "cpu")
print("使用 CUDA: {}".format(args.cuda))
# 设置随机种子以保证可重复性
def set_seed_everywhere(seed, cuda):
torch.manual_seed(seed)
if cuda:
torch.cuda.manual_seed_all(seed)
set_seed_everywhere(args.seed, args.cuda)
# 处理目录
def handle_dirs(dirpath):
if not os.path.exists(dirpath):
os.makedirs(dirpath)
handle_dirs(args.save_dir)
classifier = classifier.to(args.device)
dataset.class_weights = dataset.class_weights.to(args.device)
loss_func = nn.CrossEntropyLoss(dataset.class_weights)
optimizer = optim.Adam(classifier.parameters(), lr=args.learning_rate)
scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer=optimizer,
mode='min', factor=0.5,
patience=1)
train_state = make_train_state(args)
epoch_bar = tqdm_notebook(desc='training routine',
total=args.num_epochs,
position=0)
dataset.set_split('train')
train_bar = tqdm_notebook(desc='split=train',
total=dataset.get_num_batches(args.batch_size),
position=1,
leave=True)
dataset.set_split('val')
val_bar = tqdm_notebook(desc='split=val',
total=dataset.get_num_batches(args.batch_size),
position=1,
leave=True)
try:
for epoch_index in range(args.num_epochs):
train_state['epoch_index'] = epoch_index
# Iterate over training dataset
# setup: batch generator, set loss and acc to 0, set train mode on
dataset.set_split('train')
batch_generator = generate_batches(dataset,
batch_size=args.batch_size,
device=args.device)
running_loss = 0.0
running_acc = 0.0
classifier.train()
for batch_index, batch_dict in enumerate(batch_generator):
# the training routine is these 5 steps:
# --------------------------------------
# step 1. zero the gradients
optimizer.zero_grad()
# step 2. compute the output
y_pred = classifier(batch_dict['x_surname'])
# step 3. compute the loss
loss = loss_func(y_pred, batch_dict['y_nationality'])
loss_t = loss.item()
running_loss += (loss_t - running_loss) / (batch_index + 1)
# step 4. use loss to produce gradients
loss.backward()
# step 5. use optimizer to take gradient step
optimizer.step()
# -----------------------------------------
# compute the accuracy
acc_t = compute_accuracy(y_pred, batch_dict['y_nationality'])
running_acc += (acc_t - running_acc) / (batch_index + 1)
# update bar
train_bar.set_postfix(loss=running_loss, acc=running_acc,
epoch=epoch_index)
train_bar.update()
train_state['train_loss'].append(running_loss)
train_state['train_acc'].append(running_acc)
# Iterate over val dataset
# setup: batch generator, set loss and acc to 0; set eval mode on
dataset.set_split('val')
batch_generator = generate_batches(dataset,
batch_size=args.batch_size,
device=args.device)
running_loss = 0.
running_acc = 0.
classifier.eval()
for batch_index, batch_dict in enumerate(batch_generator):
# compute the output
y_pred = classifier(batch_dict['x_surname'])
# step 3. compute the loss
loss = loss_func(y_pred, batch_dict['y_nationality'])
loss_t = loss.to("cpu").item()
running_loss += (loss_t - running_loss) / (batch_index + 1)
# compute the accuracy
acc_t = compute_accuracy(y_pred, batch_dict['y_nationality'])
running_acc += (acc_t - running_acc) / (batch_index + 1)
val_bar.set_postfix(loss=running_loss, acc=running_acc,
epoch=epoch_index)
val_bar.update()
train_state['val_loss'].append(running_loss)
train_state['val_acc'].append(running_acc)
train_state = update_train_state(args=args, model=classifier,
train_state=train_state)
scheduler.step(train_state['val_loss'][-1])
if train_state['stop_early']:
break
train_bar.n = 0
val_bar.n = 0
epoch_bar.update()
except KeyboardInterrupt:
print("Exiting loop")
2.2.5 模型评估和预测
# 使用最佳可用模型计算测试集上的损失和精度
classifier.load_state_dict(torch.load(train_state['model_filename']))
classifier = classifier.to(args.device)
dataset.class_weights = dataset.class_weights.to(args.device)
loss_func = nn.CrossEntropyLoss(dataset.class_weights)
dataset.set_split('test')
batch_generator = generate_batches(dataset,
batch_size=args.batch_size,
device=args.device)
running_loss = 0.
running_acc = 0.
classifier.eval()
for batch_index, batch_dict in enumerate(batch_generator):
# 计算输出
y_pred = classifier(batch_dict['x_surname'])
# 计算loss
loss = loss_func(y_pred, batch_dict['y_nationality'])
loss_t = loss.item()
running_loss += (loss_t - running_loss) / (batch_index + 1)
# 计算准确率
acc_t = compute_accuracy(y_pred, batch_dict['y_nationality'])
running_acc += (acc_t - running_acc) / (batch_index + 1)
train_state['test_loss'] = running_loss
train_state['test_acc'] = running_acc
print("Test loss: {};".format(train_state['test_loss']))
print("Test Accuracy: {}".format(train_state['test_acc']))
def predict_nationality(surname, classifier, vectorizer):
"""利用新的姓氏预测国籍
Args:
surname (str): 需要分类的姓氏
classifier (SurnameClassifer): 一个现有的分类器
vectorizer (SurnameVectorizer): 相应的向量化器
Returns:
一个字典,包含可能性最大的国籍和对应概率
"""
vectorized_surname = vectorizer.vectorize(surname)
vectorized_surname = torch.tensor(vectorized_surname).view(1, -1)
result = classifier(vectorized_surname, apply_softmax=True)
probability_values, indices = result.max(dim=1)
index = indices.item()
predicted_nationality = vectorizer.nationality_vocab.lookup_index(index)
probability_value = probability_values.item()
return {'nationality': predicted_nationality, 'probability': probability_value}
new_surname = input("Enter a surname to classify: ")
classifier = classifier.to("cpu")
prediction = predict_nationality(new_surname, classifier, vectorizer)
print("{} -> {} (p={:0.2f})".format(new_surname,
prediction['nationality'],
prediction['probability']))
三、使用CNN对姓氏进行分类
3.1 卷积神经网络介绍
我们深入研究了MLPs、由一系列线性层和非线性函数构建的神经网络。mlp不是利用顺序模式的最佳工具。例如,在姓氏数据集中,姓氏可以有(不同长度的)段,这些段可以显示出相当多关于其起源国家的信息(如“O’Neill”中的“O”、“Antonopoulos”中的“opoulos”、“Nagasawa”中的“sawa”或“Zhu”中的“Zh”)。这些段的长度可以是可变的,挑战是在不显式编码的情况下捕获它们。
在本节中,我们将介绍卷积神经网络(CNN),这是一种非常适合检测空间子结构(并因此创建有意义的空间子结构)的神经网络。CNNs通过使用少量的权重来扫描输入数据张量来实现这一点。通过这种扫描,它们产生表示子结构检测(或不检测)的输出张量。
HISTORICAL CONTEXT
CNNs的名称和基本功能源于经典的数学运算卷积。卷积已经应用于各种工程学科,包括数字信号处理和计算机图形学。一般来说,卷积使用程序员指定的参数。这些参数被指定来匹配一些功能设计,如突出边缘或抑制高频声音。事实上,许多Photoshop滤镜都是应用于图像的固定卷积运算。然而,在深度学习和本实验中,我们从数据中学习卷积滤波器的参数,因此它对于解决当前的任务是最优的。
CNN Hyperparameters
为了理解不同的设计决策对CNN意味着什么,我们在图4-6中展示了一个示例。在本例中,单个“核”应用于输入矩阵。卷积运算(线性算子)的精确数学表达式对于理解这一节并不重要,但是从这个图中可以直观地看出,核是一个小的方阵,它被系统地应用于输入矩阵的不同位置。
图4-6 二维卷积运算。
输入矩阵与单个产生输出矩阵的卷积核(也称为特征映射)在输入矩阵的每个位置应用内核。在每个应用程序中,内核乘以输入矩阵的值及其自身的值,然后将这些乘法相加kernel具有以下超参数配置:kernel_size=2,stride=1,padding=0,以及dilation=1。这些超参数解释如下:虽然经典卷积是通过指定核的具体值来设计的,但是CNN是通过指定控制CNN行为的超参数来设计的,然后使用梯度下降来为给定数据集找到最佳参数。两个主要的超参数控制卷积的形状(称为kernel_size)和卷积将在输入数据张量(称为stride)中相乘的位置。还有一些额外的超参数控制输入数据张量被0填充了多少(称为padding),以及当应用到输入数据张量(称为dilation)时,乘法应该相隔多远。
3.2 CNN实现姓氏分类
3.2.1 数据集
虽然姓氏数据集之前在“示例:带有多层感知器的姓氏分类”中进行了描述,但建议参考“姓氏数据集”来了解它的描述。尽管我们使用了来自“示例:带有多层感知器的姓氏分类”中的相同数据集,但在实现上有一个不同之处:数据集由onehot向量矩阵组成,而不是一个收缩的onehot向量。为此,我们实现了一个数据集类,它跟踪最长的姓氏,并将其作为矩阵中包含的行数提供给矢量化器。列的数量是onehot向量的大小(词汇表的大小)。
3.2.2 词汇表,向量化器
在本例中,尽管词汇表和DataLoader的实现方式与“示例:带有多层感知器的姓氏分类”中的示例相同,但Vectorizer的vectorize()方法已经更改,以适应CNN模型的需要。
函数将字符串中的每个字符映射到一个整数,然后使用该整数构造一个由onehot向量组成的矩阵。重要的是,矩阵中的每一列都是不同的onehot向量。主要原因是,我们将使用的Conv1d层要求数据张量在第0维上具有批处理,在第1维上具有通道,在第2维上具有特性。
除了更改为使用onehot矩阵之外,我们还修改了矢量化器,以便计算姓氏的最大长度并将其保存为max_surname_length
class Vocabulary(object):
"""Class to process text and extract vocabulary for mapping"""
def __init__(self, token_to_idx=None, add_unk=True, unk_token="<UNK>"):
"""
Args:
token_to_idx (dict): a pre-existing map of tokens to indices
add_unk (bool): a flag that indicates whether to add the UNK token
unk_token (str): the UNK token to add into the Vocabulary
"""
if token_to_idx is None:
token_to_idx = {}
self._token_to_idx = token_to_idx
self._idx_to_token = {idx: token
for token, idx in self._token_to_idx.items()}
self._add_unk = add_unk
self._unk_token = unk_token
self.unk_index = -1
if add_unk:
self.unk_index = self.add_token(unk_token)
def to_serializable(self):
""" returns a dictionary that can be serialized """
return {'token_to_idx': self._token_to_idx,
'add_unk': self._add_unk,
'unk_token': self._unk_token}
@classmethod
def from_serializable(cls, contents):
""" instantiates the Vocabulary from a serialized dictionary """
return cls(**contents)
def add_token(self, token):
"""Update mapping dicts based on the token.
Args:
token (str): the item to add into the Vocabulary
Returns:
index (int): the integer corresponding to the token
"""
try:
index = self._token_to_idx[token]
except KeyError:
index = len(self._token_to_idx)
self._token_to_idx[token] = index
self._idx_to_token[index] = token
return index
def add_many(self, tokens):
"""Add a list of tokens into the Vocabulary
Args:
tokens (list): a list of string tokens
Returns:
indices (list): a list of indices corresponding to the tokens
"""
return [self.add_token(token) for token in tokens]
def lookup_token(self, token):
"""Retrieve the index associated with the token
or the UNK index if token isn't present.
Args:
token (str): the token to look up
Returns:
index (int): the index corresponding to the token
Notes:
`unk_index` needs to be >=0 (having been added into the Vocabulary)
for the UNK functionality
"""
if self.unk_index >= 0:
return self._token_to_idx.get(token, self.unk_index)
else:
return self._token_to_idx[token]
def lookup_index(self, index):
"""Return the token associated with the index
Args:
index (int): the index to look up
Returns:
token (str): the token corresponding to the index
Raises:
KeyError: if the index is not in the Vocabulary
"""
if index not in self._idx_to_token:
raise KeyError("the index (%d) is not in the Vocabulary" % index)
return self._idx_to_token[index]
def __str__(self):
return "<Vocabulary(size=%d)>" % len(self)
def __len__(self):
return len(self._token_to_idx)
3.2.3 分类模型
我们在本例中使用的模型是使用我们在“卷积神经网络”中介绍的方法构建的。实际上,我们在该部分中创建的用于测试卷积层的“人工”数据与姓氏数据集中使用本例中的矢量化器的数据张量的大小完全匹配。正如在示例4-19中所看到的,它与我们在“卷积神经网络”中引入的Conv1d序列既有相似之处,也有需要解释的新添加内容。具体来说,该模型类似于“卷积神经网络”,它使用一系列一维卷积来增量地计算更多的特征,从而得到一个单特征向量。
然而,本例中的新内容是使用sequence和ELU PyTorch模块。序列模块是封装线性操作序列的方便包装器。在这种情况下,我们使用它来封装Conv1d序列的应用程序。ELU是类似于实验3中介绍的ReLU的非线性函数,但是它不是将值裁剪到0以下,而是对它们求幂。ELU已经被证明是卷积层之间使用的一种很有前途的非线性(Clevert et al., 2015)。
在本例中,我们将每个卷积的通道数与num_channels超参数绑定。我们可以选择不同数量的通道分别进行卷积运算。这样做需要优化更多的超参数。我们发现256足够大,可以使模型达到合理的性能。
class SurnameClassifier(nn.Module):
def __init__(self, initial_num_channels, num_classes, num_channels):
"""
Args:
initial_num_channels (int): size of the incoming feature vector
num_classes (int): size of the output prediction vector
num_channels (int): constant channel size to use throughout network
"""
super(SurnameClassifier, self).__init__()
self.convnet = nn.Sequential(
nn.Conv1d(in_channels=initial_num_channels,
out_channels=num_channels, kernel_size=3),
nn.ELU(),
nn.Conv1d(in_channels=num_channels, out_channels=num_channels,
kernel_size=3, stride=2),
nn.ELU(),
nn.Conv1d(in_channels=num_channels, out_channels=num_channels,
kernel_size=3, stride=2),
nn.ELU(),
nn.Conv1d(in_channels=num_channels, out_channels=num_channels,
kernel_size=3),
nn.ELU()
)
self.fc = nn.Linear(num_channels, num_classes)
def forward(self, x_surname, apply_softmax=False):
"""The forward pass of the classifier
Args:
x_surname (torch.Tensor): an input data tensor.
x_surname.shape should be (batch, initial_num_channels, max_surname_length)
apply_softmax (bool): a flag for the softmax activation
should be false if used with the Cross Entropy losses
Returns:
the resulting tensor. tensor.shape should be (batch, num_classes)
"""
features = self.convnet(x_surname).squeeze(dim=2)
prediction_vector = self.fc(features)
if apply_softmax:
prediction_vector = F.softmax(prediction_vector, dim=1)
return prediction_vector
3.2.4 模型训练
args = Namespace(
# 数据和路径信息
surname_csv="data/surnames/surnames_with_splits.csv",
vectorizer_file="vectorizer.json",
model_state_file="model.pth",
save_dir="model_storage/ch4/cnn",
# 模型超参数
hidden_dim=100,
num_channels=256,
# 训练超参数
seed=1337,
learning_rate=0.001,
batch_size=128,
num_epochs=100,
early_stopping_criteria=5,
dropout_p=0.1,
# 运行时选项
cuda=False,
reload_from_files=False,
expand_filepaths_to_save_dir=True,
catch_keyboard_interrupt=True
)
if args.expand_filepaths_to_save_dir:
args.vectorizer_file = os.path.join(args.save_dir,
args.vectorizer_file)
args.model_state_file = os.path.join(args.save_dir,
args.model_state_file)
print("Expanded filepaths: ")
print("\t{}".format(args.vectorizer_file))
print("\t{}".format(args.model_state_file))
# 检查 CUDA
if not torch.cuda.is_available():
args.cuda = False
args.device = torch.device("cuda" if args.cuda else "cpu")
print("Using CUDA: {}".format(args.cuda))
# 设置随机种子
set_seed_everywhere(args.seed, args.cuda)
# 处理目录
handle_dirs(args.save_dir)
if args.reload_from_files:
# 从检查点继续训练
dataset = SurnameDataset.load_dataset_and_load_vectorizer(args.surname_csv,
args.vectorizer_file)
else:
# 创建数据集和向量化器
dataset = SurnameDataset.load_dataset_and_make_vectorizer(args.surname_csv)
dataset.save_vectorizer(args.vectorizer_file)
vectorizer = dataset.get_vectorizer()
classifier = SurnameClassifier(initial_num_channels=len(vectorizer.surname_vocab),
num_classes=len(vectorizer.nationality_vocab),
num_channels=args.num_channels)
classifer = classifier.to(args.device)
dataset.class_weights = dataset.class_weights.to(args.device)
#使用交叉熵作为loss
loss_func = nn.CrossEntropyLoss(weight=dataset.class_weights)
#使用Adam作为优化器
optimizer = optim.Adam(classifier.parameters(), lr=args.learning_rate)
#使用 ReduceLROnPlateau 调度器
scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer=optimizer,
mode='min', factor=0.5,
patience=1)
train_state = make_train_state(args)
#进度条初始化
epoch_bar = tqdm_notebook(desc='training routine',
total=args.num_epochs,
position=0)
dataset.set_split('train')
train_bar = tqdm_notebook(desc='split=train',
total=dataset.get_num_batches(args.batch_size),
position=1,
leave=True)
dataset.set_split('val')
val_bar = tqdm_notebook(desc='split=val',
total=dataset.get_num_batches(args.batch_size),
position=1,
leave=True)
#训练周期
try:
for epoch_index in range(args.num_epochs):
train_state['epoch_index'] = epoch_index
#迭代训练数据集
#setup: 批量生成器,设置损耗和acc为0,设置训练模式
dataset.set_split('train')
batch_generator = generate_batches(dataset,
batch_size=args.batch_size,
device=args.device)
running_loss = 0.0
running_acc = 0.0
classifier.train()
for batch_index, batch_dict in enumerate(batch_generator):
# 训练周期为以下5步:
# --------------------------------------
# 第一步:将梯度归零
optimizer.zero_grad()
# 第二步:计算模型输出
y_pred = classifier(batch_dict['x_surname'])
# 第三步:计算损失,并更新 running_loss
loss = loss_func(y_pred, batch_dict['y_nationality'])
loss_t = loss.item()
running_loss += (loss_t - running_loss) / (batch_index + 1)
# 第四步:反向传播计算梯度。
loss.backward()
# 第五步:优化器更新模型参数。
optimizer.step()
# -----------------------------------------
计算当前批次的准确率,并更新 running_acc。
acc_t = compute_accuracy(y_pred, batch_dict['y_nationality'])
running_acc += (acc_t - running_acc) / (batch_index + 1)
更新训练进度条,显示当前的损失、准确率和周期
train_bar.set_postfix(loss=running_loss, acc=running_acc,
epoch=epoch_index)
train_bar.update()
train_state['train_loss'].append(running_loss)
train_state['train_acc'].append(running_acc)
# 设置验证数据集
# setup: 批量生成器,设置损耗和acc为0; 设置为eval模式
dataset.set_split('val')
batch_generator = generate_batches(dataset,
batch_size=args.batch_size,
device=args.device)
running_loss = 0.
running_acc = 0.
classifier.eval()
for batch_index, batch_dict in enumerate(batch_generator):
# 计算模型输出
y_pred = classifier(batch_dict['x_surname'])
# 计算损失,并更新 running_loss
loss = loss_func(y_pred, batch_dict['y_nationality'])
loss_t = loss.item()
running_loss += (loss_t - running_loss) / (batch_index + 1)
# 计算当前批次的准确率,并更新 running_acc。
acc_t = compute_accuracy(y_pred, batch_dict['y_nationality'])
running_acc += (acc_t - running_acc) / (batch_index + 1)
val_bar.set_postfix(loss=running_loss, acc=running_acc,
epoch=epoch_index)
val_bar.update()
train_state['val_loss'].append(running_loss)
train_state['val_acc'].append(running_acc)
train_state = update_train_state(args=args, model=classifier,
train_state=train_state)
scheduler.step(train_state['val_loss'][-1])
if train_state['stop_early']:
break
train_bar.n = 0
val_bar.n = 0
epoch_bar.update()
except KeyboardInterrupt:
print("Exiting loop")
3.2.5 模型评估与预测
classifier.load_state_dict(torch.load(train_state['model_filename']))
classifier = classifier.to(args.device)
dataset.class_weights = dataset.class_weights.to(args.device)
loss_func = nn.CrossEntropyLoss(dataset.class_weights)
dataset.set_split('test')
batch_generator = generate_batches(dataset,
batch_size=args.batch_size,
device=args.device)
running_loss = 0.
running_acc = 0.
classifier.eval()
for batch_index, batch_dict in enumerate(batch_generator):
# compute the output
y_pred = classifier(batch_dict['x_surname'])
# compute the loss
loss = loss_func(y_pred, batch_dict['y_nationality'])
loss_t = loss.item()
running_loss += (loss_t - running_loss) / (batch_index + 1)
# compute the accuracy
acc_t = compute_accuracy(y_pred, batch_dict['y_nationality'])
running_acc += (acc_t - running_acc) / (batch_index + 1)
train_state['test_loss'] = running_loss
train_state['test_acc'] = running_acc
def predict_nationality(surname, classifier, vectorizer):
"""Predict the nationality from a new surname
Args:
surname (str): the surname to classifier
classifier (SurnameClassifer): an instance of the classifier
vectorizer (SurnameVectorizer): the corresponding vectorizer
Returns:
a dictionary with the most likely nationality and its probability
"""
vectorized_surname = vectorizer.vectorize(surname)
vectorized_surname = torch.tensor(vectorized_surname).unsqueeze(0)
result = classifier(vectorized_surname, apply_softmax=True)
probability_values, indices = result.max(dim=1)
index = indices.item()
predicted_nationality = vectorizer.nationality_vocab.lookup_index(index)
probability_value = probability_values.item()
return {'nationality': predicted_nationality, 'probability': probability_value}
def predict_topk_nationality(surname, classifier, vectorizer, k=5):
"""Predict the top K nationalities from a new surname
Args:
surname (str): the surname to classifier
classifier (SurnameClassifer): an instance of the classifier
vectorizer (SurnameVectorizer): the corresponding vectorizer
k (int): the number of top nationalities to return
Returns:
list of dictionaries, each dictionary is a nationality and a probability
"""
vectorized_surname = vectorizer.vectorize(surname)
vectorized_surname = torch.tensor(vectorized_surname).unsqueeze(dim=0)
prediction_vector = classifier(vectorized_surname, apply_softmax=True)
probability_values, indices = torch.topk(prediction_vector, k=k)
# returned size is 1,k
probability_values = probability_values[0].detach().numpy()
indices = indices[0].detach().numpy()
results = []
for kth_index in range(k):
nationality = vectorizer.nationality_vocab.lookup_index(indices[kth_index])
probability_value = probability_values[kth_index]
results.append({'nationality': nationality,
'probability': probability_value})
return results
new_surname = input("Enter a surname to classify: ")
k = int(input("How many of the top predictions to see? "))
if k > len(vectorizer.nationality_vocab):
print("Sorry! That's more than the # of nationalities we have.. defaulting you to max size :)")
k = len(vectorizer.nationality_vocab)
predictions = predict_topk_nationality(new_surname, classifier, vectorizer, k=k)
print("Top {} predictions:".format(k))
print("===================")
for prediction in predictions:
print("{} -> {} (p={:0.2f})".format(new_surname,
prediction['nationality'],
prediction['probability']))
四、总结
在利用机器学习方法进行姓氏分类的任务中,多层感知机和卷积神经网络(是两种常被考虑的模型。以下是我基于这两种模型应用于姓氏分类任务的一个简短总结:
多层感知机(MLP)应用于姓氏分类:
-
基本原理:多层感知机是一种前馈神经网络,通过一系列线性变换和非线性激活函数处理输入数据,适用于处理结构化数据。在姓氏分类任务中,MLP可以直接接受经过编码处理的文本特征(如one-hot编码或词向量)作为输入。
-
优势:
- 相对简单,容易理解和实现。
- 可以处理大小固定的输入,对于一些简单的文本分类任务可能已经足够。
-
局限性:
- 对于文本这类具有天然序列特性和局部依赖性的数据,MLP缺乏有效捕捉这些特征的机制,可能不如专门设计用于处理此类数据的模型高效。
- 需要较大的特征工程来提取有用的文本特征,如手工设计的特征或高质量的预训练词向量。
卷积神经网络(CNN)应用于姓氏分类:
-
基本原理:CNN最初为图像识别设计,但其局部连接和权值共享的特性同样适用于文本数据。在处理文本时,通过将文本视为一维“图像”(即字符或词的序列),CNN能够通过卷积层捕获局部特征和模式。
-
优势:
- 局部相关性捕捉:CNN的卷积核能有效地捕获文本中的n-gram特征,这对于识别某些特定的字符组合或模式很有帮助,尤其是在姓氏这种长度较短且结构相对固定的文本分类中。
- 参数共享:减少了模型参数的数量,有助于防止过拟合,同时加快了训练速度。
-
局限性:
- 尽管CNN能够捕捉局部特征,但在处理需要理解长距离依赖关系的复杂文本分类任务时,可能不如循环神经网络(RNN)或Transformer等模型效果好。
- 对于非常短的文本如单一的姓氏,CNN的有效性可能会受限,因为局部结构的区分度可能不如在更长文本中那么明显。
总结:
在姓氏分类任务中,MLP和CNN各有优劣。MLP简单易行,适合快速原型设计;而CNN能够更好地捕捉文本中的局部特征模式,尤其适合处理具有一定结构特征的文本分类问题。选择哪种模型取决于任务的具体需求、数据的性质以及计算资源的限制。实际应用中,也可以尝试结合多种模型或采用更先进的深度学习架构,以进一步提升分类性能。