Pytorch基礎 logistic回歸

import torch
import torch.nn as nn
import numpy as np
print
(torch.__version__)

1.1.0

logistic回歸是一中廣義線性回歸，與多重線性回歸分析有很多相同之處。它們模型形式上基本相同，都具有wx+b,其中w和b是待求解引數，區別在於因變數不同，多重線性回歸直接將wx+b作為因變數，即y=wx+b，而logistic回歸則通過函式l將wx+b對應乙個隱狀態p，p=l(wx+b)，根據p與1-p的大小決定因變數的值。

l為logistic函式時為logistic回歸，l為多項式函式時為多項式回歸

logistic回歸主要進行二分類**：sigmoid函式就是常見的logistic函式，因為sigmoid函式的輸出時0~1之間的概率，當概率大於0.5時**為1，小於0.5時為0

# 載入資料
data = np.loadtxt(
'./data/german.data-numeric'
)n, l = data.shape
# 資料歸一化
for j in
range
(l -1)
: meanval = np.mean(data[
:, j]
) stdval = np.std(data[
:, j]
) data[
:, j]
=(data[
:, j]
- meanval)
/ stdval

# 打亂資料
np.random.shuffle(data)

# 劃分訓練集和測試集
train_data = data[
:900
,:l -1]
train_tag = data[
:900
, l -1]
-1test_data = data[
900:
,:l -1]
test_tag = data[
900:
, l -1]
-1

# 定義網路
class
net(nn.module)
:def
__init__
(self)
:super
(net, self)
.__init__(
) self.fc = nn.linear(24,
2)self.sigmoid = nn.sigmoid(
)def
forward
(self, x)
: out = self.fc(x)
out = self.sigmoid(out)
return out

def
test
(pred, lab)
: t = pred.
max(-1
)[1]
== lab
return torch.mean(t.
float()
)

net = net(
)critertion = nn.crossentropyloss(
)# 使用交叉熵損失
optimizer = torch.optim.adam(net.parameters())
# adam優化器
epochs =
1000
for epoch in
range
(epochs)
: net.train(
)# 指定模型為訓練模型，計算梯度
x = torch.from_numpy(train_data)
.float()
y = torch.from_numpy(train_tag)
.long()
y_pred = net(x)
loss = critertion(y_pred, y)
# 計算損失
optimizer.zero_grad(
)# 權重置零
loss.backward(
)# 反向傳播
optimizer.step()if
(epoch +1)
%100==0
: net.
eval()
#指定模型計算模式
test_in = torch.from_numpy(test_data)
.float()
test_t = torch.from_numpy(test_tag)
.long()
test_out = net(test_in)
# 使用測試函式計算準確率
accu = test(test_out, test_t)
print
('epoch: {}, loss: {}, accuracy:{}'
.format
( epoch +
1, loss.item(
), accu)
)

epoch: 100, loss: 0.6658604145050049, accuracy:0.699999988079071 epoch: 200, loss: 0.6306068897247314, accuracy:0.8199999928474426 epoch: 300, loss: 0.6095178723335266, accuracy:0.8100000023841858 epoch: 400, loss: 0.5955496430397034, accuracy:0.8100000023841858 epoch: 500, loss: 0.5853410363197327, accuracy:0.800000011920929 epoch: 600, loss: 0.5774123072624207, accuracy:0.8199999928474426 epoch: 700, loss: 0.5710282921791077, accuracy:0.8199999928474426 epoch: 800, loss: 0.5657661557197571, accuracy:0.8199999928474426 epoch: 900, loss: 0.5613517165184021, accuracy:0.8199999928474426

epoch: 1000, loss: 0.5575944185256958, accuracy:0.8199999928474426

Pytorch基礎 logistic回歸

邏輯回歸之 Logist 推導

pytorch基礎函式

Pytorch基礎操作

Pytorch基礎 logistic回歸

邏輯回歸 之 Logist 推導

pytorch基礎函式

Pytorch基礎操作

相關推薦

邏輯回歸之 Logist 推導