task3 task4 and task5
1.过拟合、欠拟合及其解决方案
解决方法包括:验证数据集和交叉验证
权重衰减
L2 范数正则化(regularization)
例如在线性回归中加入带有l2范数惩罚项的损失函数。
当 λ 较大时,惩罚项在损失函数中的比重较大,这通常会使学到的权重参数的元素较接近0。当 λ 设为0时,惩罚项完全不起作用。
2.(1)梯度消失、梯度爆炸以及Kaggle房价预测
当神经网络的层数较多时,模型的数值稳定性容易变差
(2)考虑环境因素
协变量偏移
标签偏移
概念偏移
Kaggle 房价预测实战
%matplotlib inline
import torch
import torch.nn as nn
import numpy as np
import pandas as pd
import sys
sys.path.append("/home/kesci/input")
import d2lzh1981 as d2l
print(torch.__version__)
torch.set_default_tensor_type(torch.FloatTensor)
获取和读取数据集
test_data = pd.read_csv("/home/kesci/input/houseprices2807/house-prices-advanced-regression-techniques/test.csv")
train_data = pd.read_csv("/home/kesci/input/houseprices2807/house-prices-advanced-regression-techniques/train.csv")
预处理数据
numeric_features = all_features.dtypes[all_features.dtypes != 'object'].index
all_features[numeric_features] = all_features[numeric_features].apply(
lambda x: (x - x.mean()) / (x.std()))
# 标准化后,每个数值特征的均值变为0,所以可以直接用0来替换缺失值
all_features[numeric_features] = all_features[numeric_features].fillna(0)
# dummy_na=True将缺失值也当作合法的特征值并为其创建指示特征
all_features = pd.get_dummies(all_features, dummy_na=True)
all_features.shape
#转为tersor
n_train = train_data.shape[0]
train_features = torch.tensor(all_features[:n_train].values, dtype=torch.float)
test_features = torch.tensor(all_features[n_train:].values, dtype=torch.float)
train_labels = torch.tensor(train_data.SalePrice.values, dtype=torch.float).view(-1, 1)
训练模型
loss = torch.nn.MSELoss()
def get_net(feature_num):
net = nn.Linear(feature_num, 1)
for param in net.parameters():
nn.init.normal_(param, mean=0, std=0.01)
return net
def train(net, train_features, train_labels, test_features, test_labels,
num_epochs, learning_rate, weight_decay, batch_size):
train_ls, test_ls = [], []
dataset = torch.utils.data.TensorDataset(train_features, train_labels)
train_iter = torch.utils.data.DataLoader(dataset, batch_size, shuffle=True)
# 这里使用了Adam优化算法
optimizer = torch.optim.Adam(params=net.parameters(), lr=learning_rate, weight_decay=weight_decay)
net = net.float()
for epoch in range(num_epochs):
for X, y in train_iter:
l = loss(net(X.float()), y.float())
optimizer.zero_grad()
l.backward()
optimizer.step()
train_ls.append(log_rmse(net, train_features, train_labels))
if test_labels is not None:
test_ls.append(log_rmse(net, test_features, test_labels))
return train_ls, test_ls
我们在模型选择、欠拟合和过拟合中介绍了 K 折交叉验证。它将被用来选择模型设计并调节超参数
预测函数
def train_and_pred(train_features, test_features, train_labels, test_data,
num_epochs, lr, weight_decay, batch_size):
net = get_net(train_features.shape[1])
train_ls, _ = train(net, train_features, train_labels, None, None,
num_epochs, lr, weight_decay, batch_size)
d2l.semilogy(range(1, num_epochs + 1), train_ls, 'epochs', 'rmse')
print('train rmse %f' % train_ls[-1])
preds = net(test_features).detach().numpy()
test_data['SalePrice'] = pd.Series(preds.reshape(1, -1)[0])
submission = pd.concat([test_data['Id'], test_data['SalePrice']], axis=1)
submission.to_csv('./submission.csv', index=False)
# sample_submission_data = pd.read_csv("../input/house-prices-advanced-regression-techniques/sample_submission.csv")
train_and_pred(train_features, test_features, train_labels, test_data, num_epochs, lr, weight_decay, batch_size)
3.循环神经网络进阶
RNN存在的问题:梯度较容易出现衰减或爆炸(BPTT)
⻔控循环神经⽹络:捕捉时间序列中时间步距离较⼤的依赖关系
根据相关 的gru模型公式可写出gru模型
def gru(inputs, state, params):
W_xz, W_hz, b_z, W_xr, W_hr, b_r, W_xh, W_hh, b_h, W_hq, b_q = params
H, = state
outputs = []
for X in inputs:
Z = torch.sigmoid(torch.matmul(X, W_xz) + torch.matmul(H, W_hz) + b_z)
R = torch.sigmoid(torch.matmul(X, W_xr) + torch.matmul(H, W_hr) + b_r)
H_tilda = torch.tanh(torch.matmul(X, W_xh) + R * torch.matmul(H, W_hh) + b_h)
H = Z * H + (1 - Z) * H_tilda
Y = torch.matmul(H, W_hq) + b_q
outputs.append(Y)
return outputs, (H,)
LSTM
长短期记忆long short-term memory :
遗忘门:控制上一时间步的记忆细胞 输入门:控制当前时间步的输入
输出门:控制从记忆细胞到隐藏状态
记忆细胞:⼀种特殊的隐藏状态的信息的流动
LSTM模型
def lstm(inputs, state, params):
[W_xi, W_hi, b_i, W_xf, W_hf, b_f, W_xo, W_ho, b_o, W_xc, W_hc, b_c, W_hq, b_q] = params
(H, C) = state
outputs = []
for X in inputs:
I = torch.sigmoid(torch.matmul(X, W_xi) + torch.matmul(H, W_hi) + b_i)
F = torch.sigmoid(torch.matmul(X, W_xf) + torch.matmul(H, W_hf) + b_f)
O = torch.sigmoid(torch.matmul(X, W_xo) + torch.matmul(H, W_ho) + b_o)
C_tilda = torch.tanh(torch.matmul(X, W_xc) + torch.matmul(H, W_hc) + b_c)
C = F * C + I * C_tilda
H = O * C.tanh()
Y = torch.matmul(H, W_hq) + b_q
outputs.append(Y)
return outputs, (H, C)
task4
1.机器翻译及相关技术
机器翻译(MT):将一段文本从一种语言自动翻译为另一种语言,用神经网络解决这个问题通常称为神经机器翻译(NMT)。 主要特征:输出是单词序列而不是单个单词。 输出序列的长度可能与源序列的长度不同。
Beam Search
2.注意力机制与Seq2seq模型
注意力机制框架
Attention 是一种通用的带权池化方法,输入由两部分构成:询问(query)和键值对(key-value pairs)。 ki∈Rdk,vi∈Rdv . Query q∈Rdq , attention layer得到输出与value的维度一致 o∈Rdv . 对于一个query来说,attention layer 会与每一个key计算注意力分数并进行权重的归一化,输出的向量 o 则是value的加权求和,而每个key计算的权重与value一一对应。
为了计算输出,我们首先假设有一个函数 α 用于计算query和key的相似性,然后可以计算所有的 attention scores a1,…,an by
ai=α(q,ki).
我们使用 softmax函数 获得注意力权重:
b1,…,bn=softmax(a1,…,an).
最终的输出就是value的加权求和:
o=∑i=1nbivi.
点积注意力
多层感知机注意力
注意力层显式地选择相关的信息。
注意层的内存由键-值对组成,因此它的输出接近于键类似于查询的值。
seq2seq训练
import zipfile
import torch
import requests
from io import BytesIO
from torch.utils import data
import sys
import collections
class Vocab(object): # This class is saved in d2l.
def __init__(self, tokens, min_freq=0, use_special_tokens=False):
# sort by frequency and token
counter = collections.Counter(tokens)
token_freqs = sorted(counter.items(), key=lambda x: x[0])
token_freqs.sort(key=lambda x: x[1], reverse=True)
if use_special_tokens:
# padding, begin of sentence, end of sentence, unknown
self.pad, self.bos, self.eos, self.unk = (0, 1, 2, 3)
tokens = ['', '', '', '']
else:
self.unk = 0
tokens = ['']
tokens += [token for token, freq in token_freqs if freq >= min_freq]
self.idx_to_token = []
self.token_to_idx = dict()
for token in tokens:
self.idx_to_token.append(token)
self.token_to_idx[token] = len(self.idx_to_token) - 1
def __len__(self):
return len(self.idx_to_token)
def __getitem__(self, tokens):
if not isinstance(tokens, (list, tuple)):
return self.token_to_idx.get(tokens, self.unk)
else:
return [self.__getitem__(token) for token in tokens]
def to_tokens(self, indices):
if not isinstance(indices, (list, tuple)):
return self.idx_to_token[indices]
else:
return [self.idx_to_token[index] for index in indices]
def load_data_nmt(batch_size, max_len, num_examples=1000):
"""Download an NMT dataset, return its vocabulary and data iterator."""
# Download and preprocess
def preprocess_raw(text):
text = text.replace('\u202f', ' ').replace('\xa0', ' ')
out = ''
for i, char in enumerate(text.lower()):
if char in (',', '!', '.') and text[i-1] != ' ':
out += ' '
out += char
return out
with open('/home/kesci/input/fraeng6506/fra.txt', 'r') as f:
raw_text = f.read()
text = preprocess_raw(raw_text)
# Tokenize
source, target = [], []
for i, line in enumerate(text.split('\n')):
if i >= num_examples:
break
parts = line.split('\t')
if len(parts) >= 2:
source.append(parts[0].split(' '))
target.append(parts[1].split(' '))
# Build vocab
def build_vocab(tokens):
tokens = [token for line in tokens for token in line]
return Vocab(tokens, min_freq=3, use_special_tokens=True)
src_vocab, tgt_vocab = build_vocab(source), build_vocab(target)
# Convert to index arrays
def pad(line, max_len, padding_token):
if len(line) > max_len:
return line[:max_len]
return line + [padding_token] * (max_len - len(line))
def build_array(lines, vocab, max_len, is_source):
lines = [vocab[line] for line in lines]
if not is_source:
lines = [[vocab.bos] + line + [vocab.eos] for line in lines]
array = torch.tensor([pad(line, max_len, vocab.pad) for line in lines])
valid_len = (array != vocab.pad).sum(1)
return array, valid_len
src_vocab, tgt_vocab = build_vocab(source), build_vocab(target)
src_array, src_valid_len = build_array(source, src_vocab, max_len, True)
tgt_array, tgt_valid_len = build_array(target, tgt_vocab, max_len, False)
train_data = data.TensorDataset(src_array, src_valid_len, tgt_array, tgt_valid_len)
train_iter = data.DataLoader(train_data, batch_size, shuffle=True)
return src_vocab, tgt_vocab, train_iter
embed_size, num_hiddens, num_layers, dropout = 32, 32, 2, 0.0
batch_size, num_steps = 64, 10
lr, num_epochs, ctx = 0.005, 500, d2l.try_gpu()
src_vocab, tgt_vocab, train_iter = load_data_nmt(batch_size, num_steps)
encoder = d2l.Seq2SeqEncoder(
len(src_vocab), embed_size, num_hiddens, num_layers, dropout)
decoder = Seq2SeqAttentionDecoder(
len(tgt_vocab), embed_size, num_hiddens, num_layers, dropout)
model = d2l.EncoderDecoder(encoder, decoder)
3.Transformer
整合CNN和RNN的优势
多头注意力层
自注意力模型是一个正规的注意力模型,序列的每一个元素对应的key,value,query是完全一致的
头注意力层包含 h 个并行的自注意力层,每一个这种层被成为一个head。对每个头来说,在进行注意力计算之前,我们会将query、key和value用三个现行层进行映射,这 h 个注意力头的输出将会被拼接之后输入最后一个线性层进行整合。
task5
1.卷积神经网络基础(卷积层,池化层,全连接层)
填充
填充(padding)是指在输入高和宽的两侧填充元素(通常是0元素),图2里我们在原输入高和宽的两侧分别添加了值为0的元素。
步幅
在互相关运算中,卷积核在输入数组上滑动,每次滑动的行数与列数即是步幅(stride)。此前我们使用的步幅都是1,图3展示了在高上步幅为3、在宽上步幅为2的二维互相关运算。
多输入通道和多输出通道
卷积层与全连接层的对比
二维卷积层经常用于处理图像,与此前的全连接层相比,它主要有两个优势:
一是全连接层把图像展平成一个向量,在输入图像上相邻的元素可能因为展平操作不再相邻,网络难以捕捉局部信息。而卷积层的设计,天然地具有提取局部信息的能力。
二是卷积层的参数量更少。不考虑偏置的情况下,一个形状为 (ci,co,h,w) 的卷积核的参数量是 ci×co×h×w ,与输入图像的宽高无关。假如一个卷积层的输入和输出形状分别是 (c1,h1,w1) 和 (c2,h2,w2) ,如果要用全连接层进行连接,参数数量就是 c1×c2×h1×w1×h2×w2 。使用卷积层可以以较少的参数数量来处理更大的图像。
池化
二维池化层
池化层主要用于缓解卷积层对位置的过度敏感性。同卷积层一样,池化层每次对输入数据的一个固定形状窗口(又称池化窗口)中的元素计算输出,池化层直接计算池化窗口内元素的最大值或者平均值,该运算也分别叫做最大池化或平均池化。图6展示了池化窗口形状为 2×2 的最大池化。
2.LeNet
LeNet分为卷积层块和全连接层块两个部分
#训练函数
def train_ch5(net, train_iter, test_iter,criterion, num_epochs, batch_size, device,lr=None):
"""Train and evaluate a model with CPU or GPU."""
print('training on', device)
net.to(device)
optimizer = optim.SGD(net.parameters(), lr=lr)
for epoch in range(num_epochs):
train_l_sum = torch.tensor([0.0],dtype=torch.float32,device=device)
train_acc_sum = torch.tensor([0.0],dtype=torch.float32,device=device)
n, start = 0, time.time()
for X, y in train_iter:
net.train()
optimizer.zero_grad()
X,y = X.to(device),y.to(device)
y_hat = net(X)
loss = criterion(y_hat, y)
loss.backward()
optimizer.step()
with torch.no_grad():
y = y.long()
train_l_sum += loss.float()
train_acc_sum += (torch.sum((torch.argmax(y_hat, dim=1) == y))).float()
n += y.shape[0]
test_acc = evaluate_accuracy(test_iter, net,device)
print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f, '
'time %.1f sec'
% (epoch + 1, train_l_sum/n, train_acc_sum/n, test_acc,
time.time() - start))
3.卷积神经网络进阶
import time
import torch
from torch import nn, optim
import torchvision
import numpy as np
import sys
sys.path.append("/home/kesci/input/")
import d2lzh1981 as d2l
import os
import torch.nn.functional as F
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
class AlexNet(nn.Module):
def __init__(self):
super(AlexNet, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(1, 96, 11, 4), # in_channels, out_channels, kernel_size, stride, padding
nn.ReLU(),
nn.MaxPool2d(3, 2), # kernel_size, stride
# 减小卷积窗口,使用填充为2来使得输入与输出的高和宽一致,且增大输出通道数
nn.Conv2d(96, 256, 5, 1, 2),
nn.ReLU(),
nn.MaxPool2d(3, 2),
# 连续3个卷积层,且使用更小的卷积窗口。除了最后的卷积层外,进一步增大了输出通道数。
# 前两个卷积层后不使用池化层来减小输入的高和宽
nn.Conv2d(256, 384, 3, 1, 1),
nn.ReLU(),
nn.Conv2d(384, 384, 3, 1, 1),
nn.ReLU(),
nn.Conv2d(384, 256, 3, 1, 1),
nn.ReLU(),
nn.MaxPool2d(3, 2)
)
# 这里全连接层的输出个数比LeNet中的大数倍。使用丢弃层来缓解过拟合
self.fc = nn.Sequential(
nn.Linear(256*5*5, 4096),
nn.ReLU(),
nn.Dropout(0.5),
#由于使用CPU镜像,精简网络,若为GPU镜像可添加该层
#nn.Linear(4096, 4096),
#nn.ReLU(),
#nn.Dropout(0.5),
# 输出层。由于这里使用Fashion-MNIST,所以用类别数为10,而非论文中的1000
nn.Linear(4096, 10),
)
def forward(self, img):
feature = self.conv(img)
output = self.fc(feature.view(img.shape[0], -1))
return output
使用重复元素的网络(VGG)
VGG:通过重复使⽤简单的基础块来构建深度模型。
Block:数个相同的填充为1、窗口形状为 3×3 的卷积层,接上一个步幅为2、窗口形状为 2×2 的最大池化层。
卷积层保持输入的高和宽不变,而池化层则对其减半
关键代码
def vgg(conv_arch, fc_features, fc_hidden_units=4096):
net = nn.Sequential()
# 卷积层部分
for i, (num_convs, in_channels, out_channels) in enumerate(conv_arch):
# 每经过一个vgg_block都会使宽高减半
net.add_module("vgg_block_" + str(i+1), vgg_block(num_convs, in_channels, out_channels))
# 全连接层部分
net.add_module("fc", nn.Sequential(d2l.FlattenLayer(),
nn.Linear(fc_features, fc_hidden_units),
nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(fc_hidden_units, fc_hidden_units),
nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(fc_hidden_units, 10)
))
return net
⽹络中的⽹络(NiN)
LeNet、AlexNet和VGG:先以由卷积层构成的模块充分抽取 空间特征,再以由全连接层构成的模块来输出分类结果。
NiN:串联多个由卷积层和“全连接”层构成的小⽹络来构建⼀个深层⽹络。
⽤了输出通道数等于标签类别数的NiN块,然后使⽤全局平均池化层对每个通道中所有元素求平均并直接⽤于分类
class GlobalAvgPool2d(nn.Module):
# 全局平均池化层可通过将池化窗口形状设置成输入的高和宽实现
def __init__(self):
super(GlobalAvgPool2d, self).__init__()
def forward(self, x):
return F.avg_pool2d(x, kernel_size=x.size()[2:])
net = nn.Sequential(
nin_block(1, 96, kernel_size=11, stride=4, padding=0),
nn.MaxPool2d(kernel_size=3, stride=2),
nin_block(96, 256, kernel_size=5, stride=1, padding=2),
nn.MaxPool2d(kernel_size=3, stride=2),
nin_block(256, 384, kernel_size=3, stride=1, padding=1),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.Dropout(0.5),
# 标签类别数是10
nin_block(384, 10, kernel_size=3, stride=1, padding=1),
GlobalAvgPool2d(),
# 将四维的输出转成二维的输出,其形状为(批量大小, 10)
d2l.FlattenLayer())