Linear Regression
用mxnet写了最简单的线性回归
代码
from mxnet
import autograd
, nd
from mxnet
.gluon
import data
as gdata
from mxnet
.gluon
import nn
from mxnet
import init
from mxnet
.gluon
import loss
as gloss
from mxnet
import gluon
num_inputs
= 2
num_examples
= 1000
true_w
= [2.1, -5]
true_b
= 3.4
samples
= nd
.random
.normal
(scale
=1, shape
=(num_examples
, num_inputs
))
labels
= true_w
[0] * samples
[:, 0] + true_w
[1] * samples
[:, 1] + true_b
labels
+= nd
.random
.normal
(scale
=0.01, shape
=labels
.shape
)
batch_size
= 10
dataset
= gdata
.ArrayDataset
(samples
, labels
)
data_iter
= gdata
.DataLoader
(dataset
, batch_size
, shuffle
=True)
net
= nn
.Sequential
()
net
.add
(nn
.Dense
(1))
net
.initialize
((init
.Normal
(sigma
=0.01)))
loss
= gloss
.L2Loss
()
trainer
= gluon
.Trainer
(net
.collect_params
(), 'sgd', {'learning_rate': 0.03})
num_epochs
= 5
for epoch
in range(1, num_epochs
+ 1):
for X
, y
in data_iter
:
with autograd
.record
():
l
= loss
(net
(X
), y
)
l
.backward
()
trainer
.step
(batch_size
)
l
= loss
(net
(samples
), labels
)
print('epoch {}, loss: {}'.format(epoch
, l
.mean
().asnumpy
()))
dense
= net
[0]
print('真实权重:{},训练权重:{}'.format(true_w
, dense
.weight
.data
()))
print('真是偏差:{},训练偏差:{}'.format(true_b
, dense
.bias
.data
()))
结果
epoch 1, loss: [0.04274421] epoch 2, loss: [0.00014196] epoch 3, loss: [4.8662067e-05] epoch 4, loss: [4.8426606e-05] epoch 5, loss: [4.8332287e-05] 真实权重:[2.1, -5],训练权重: [[ 2.099422 -5.0002365]] <NDArray 1x2 @cpu(0)> 真是偏差:3.4,训练偏差: [3.4005916] <NDArray 1 @cpu(0)>
