1 Intro to Deep Learning

本周总结：

机器学习——定义期望的输出，然后让我们的算法自己学习到到达那里的步骤
3种学习方式——监督学习，无监督学习，强化学习
线性回归模型显示了变量间的独立性关系，创造最佳拟合的线

第一周基础没什么好说的，编个小程序热热身。

实例

import pandas as pd
from sklearn import linear_model
import matplotlib.pyplot as plt

# read data
dataframe = pd.read_fwf('brain_body.txt')
x_values = dataframe[['Brain']]
y_values = dataframe[['Body']]

#train model on data
body_reg = linear_model.LinearRegression()
body_reg.fit(x_values, y_values)

#visualize reasults
plt.scatter(x_values, y_values)
plt.plot(x_values, body_reg.predict(x_values))
plt.show()

Challenge

The challenge for this video is to use scikit-learn to create a line of best fit for the included ‘challenge_dataset’. Then, make a prediction for an existing data point and see how close it matches up to the actual value. Print out the error you get. You can use scikit-learn’s documentation for more help.

Bonus points if you perform linear regression on a dataset with 3 different variables

# Mine

import pandas as pd
from sklearn import linear_model
import matplotlib.pyplot as plt

# read data
df = pd.read_csv('challenge_dataset.txt', names=['X','Y'])
x_values = df[['X']]
y_values = df[['Y']]

# train model on data
body_reg = linear_model.LinearRegression()
body_reg.fit(x_values, y_values)

# visualize reasults
plt.scatter(x_values, y_values)
plt.plot(x_values, body_reg.predict(x_values))
plt.show()

# average error
np_data = np.array(df['Y'])
np.mean(np.abs(np_data - body_reg.predict(x_values)))

Ludo’s winning code:

%matplotlib inline

# Imports
import matplotlib.pyplot as plt
import matplotlib as mpl
import pandas as pd
import seaborn as sns
import sklearn
import numpy as np

Regression with 2 variables

1. Import and visualize data

df = pd.read_csv('challenge_dataset.txt', names=['X','Y'])

2. 2D Regression

from sklearn.model_selection import train_test_split

# 随机划分训练集和测试集
X_train, X_test, y_train, y_test = np.asarray(train_test_split(df['X'], df['Y'], test_size=0.1))

from sklearn.linear_model import LinearRegression

# 创建线性回归模型，训练模型
reg = LinearRegression()
reg.fit(X_train.values.reshape(-1,1), y_train.values.reshape(-1,1))

# 测试集测试
print('Score: ', reg.score(X_test.values.reshape(-1,1), y_test.values.reshape(-1,1)))

3. Plot regression and visualize results

x_line = np.arange(5,25).reshape(-1,1)
sns.regplot(x=df['X'], y=df['Y'], data=df, fit_reg=False)
plt.plot(x_line, reg.predict(x_line))
plt.show()

详细代码见：https://github.com/ludobouan/linear-regression-sklearn

小结

这系列深度学习课程真是挺不错的，每天听一节，自己做一下挑战题，再看看每周榜首的代码。一边学习常见的几个库，一边再巩固Python基础语法，一个假期应该会提示不少的。