File:Traintest.svg
Summary
Description |
English: Plots showing a training set and a test set from the same statistical population. Two curves are fit to the training set, one of which is an overfit. By plotting these curves with the test data, the overfitting can be seen. |
Date | |
Source | Own work |
Author | Skbkekas |
Other versions |
[ ]
|
SVG development | |
Source code | Python codeimport numpy as np
import matplotlib.pyplot as plt
m = 0.2 ## mesh on the abscissa
s = 3 ## standard deviation of errors
def pdesign(X, d):
"""Generate a polynomial design matrix on X of order d."""
V = X[:,np.newaxis]
F = [V**k for k in range(d+1)]
D = np.concatenate(F, axis=1)
return D
def regfit(Y, D):
"""Regress Y on D using least squares."""
U,S,Vt = np.linalg.svd(D,0)
V = np.transpose(Vt)
return np.dot(U, np.dot(np.transpose(U), Y))
X = np.arange(-2, 2, m, dtype=np.float64)
D1 = pdesign(X, 3)
D2 = pdesign(X, 13)
EY = X + X**3
Y1 = EY + np.random.normal(size=len(X))*s
Y2 = EY + np.random.normal(size=len(X))*s
Yhat1 = regfit(Y1, D1)
Yhat2 = regfit(Y1, D2)
plt.clf()
plt.figure(figsize=(8,3))
ax1 = plt.axes([0.06,0.1,0.4,0.8])
plt.title("Training set")
plt.plot(X, Y1, 'o')
plt.hold(True)
plt.plot(X, Yhat1, '-', color='green')
plt.plot(X, Yhat2, '-', color='orange')
ax1.set_ylim(-10, 10)
ax1.set_xticks([-2,-1,0,1,2])
ax2 = plt.axes([0.56,0.1,0.4,0.8])
plt.title("Test set")
plt.plot(X, Y2, 'o')
plt.plot(X, Yhat1, '-', color='green')
plt.plot(X, Yhat2, '-', color='orange')
ax2.set_xticks([-2,-1,0,1,2])
ax2.set_ylim(-10, 10)
plt.savefig("traintest.png")
plt.savefig("traintest.svg")
print ((Yhat1-Y1)**2).mean()
print ((Yhat2-Y1)**2).mean()
print ((Yhat1-Y2)**2).mean()
print ((Yhat2-Y2)**2).mean()
|
Licensing
I, the copyright holder of this work, hereby publish it under the following license:
This file is licensed under the Creative Commons Attribution 3.0 Unported license.
- You are free:
- to share – to copy, distribute and transmit the work
- to remix – to adapt the work
- Under the following conditions:
- attribution – You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.