Plotting multiple curve line plots in python with Matplotlib

Question

I am trying to plot 3 loss curves in one figure. I have the following problems.

I can't get a smooth curve, instead it joins point-to-point in a sharp line
How do I change the scales of the axis, to display the MSE loss which disappears because it is too small?

from matplotlib import pyplot as plt

epochs=list(range(5000,50001,5000))
print(epochs)
mae_loss=[0.500225365,
0.000221096,
0.000060971,
0.000060323,
0.000059905,
0.000059579,
0.000059274,
0.000058972,
0.000058697,
0.000058476]

mse_loss=[0.135419831,
0.018331185,
0.002481434,
0.000335913,
0.000045486,
0.000006180,
0.000000867,
0.000000147,
0.000000042,
0.000000042]

rmse_loss=[0.500225306,
0.000293739,
0.000126985,
0.000121944,
0.000119484,
0.000117791,
0.000116400,
0.000115198,
0.000114148,
0.000113228]

plt.plot(epochs, mae_loss, 'b', label='MAE')
plt.plot(epochs, mse_loss, 'r', label='MSE')
plt.plot(epochs, mse_loss, 'g', label='RMSE')
plt.legend()
plt.show()

Sheldore · Accepted Answer · 2019-05-31 22:07:15Z

3

You would require some interpolation methods to get a smooth spline/curve. That is in itself a different question. I will answer the question regarding the different scales. Since the order of magnitudes of your data are quite different, in such situations, the best solution is to use a logarithmic y-scale using semilogy. P.S: You had written mse_loss instead of rmse_loss in the last plot line.

plt.semilogy(epochs, mae_loss, 'b', label='MAE')
plt.semilogy(epochs, mse_loss, 'r', label='MSE')
plt.semilogy(epochs, rmse_loss, 'g', label='RMSE')
plt.legend()
plt.show()

answered May 31, 2019 at 22:07

Sheldore

39.2k9 gold badges63 silver badges76 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Yohst · Accepted Answer · 2019-05-31 22:21:54Z

1

To smooth your plots:

import matplotlib.pyplot as plt
import numpy as np
from scipy.interpolate import make_interp_spline, BSpline

def create_spline_from(x, y, resolution):
  new_x = np.linspace(x[0], x[-1], resolution)
  y_spline = make_interp_spline(x, y, k=3)
  new_y= y_spline(new_x)

  return (new_x, new_y)

epochs=list(range(5000,50001,5000))
print(epochs)
mae_loss=[0.500225365,
0.000221096,
0.000060971,
0.000060323,
0.000059905,
0.000059579,
0.000059274,
0.000058972,
0.000058697,
0.000058476]

mse_loss=[0.135419831,
0.018331185,
0.002481434,
0.000335913,
0.000045486,
0.000006180,
0.000000867,
0.000000147,
0.000000042,
0.000000042]

rmse_loss=[0.500225306,
0.000293739,
0.000126985,
0.000121944,
0.000119484,
0.000117791,
0.000116400,
0.000115198,
0.000114148,
0.000113228]

x, y = create_spline_from(epochs, mae_loss, 50)
plt.plot(x, y, 'b', label='MAE')

x, y = create_spline_from(epochs, mse_loss, 50)
plt.plot(x, y, 'r', label='MSE')

x, y = create_spline_from(epochs, rmse_loss, 50)
plt.plot(x, y, 'g', label='RMSE')
plt.legend()
plt.show()

answered May 31, 2019 at 22:21

Yohst

1,95021 silver badges42 bronze badges

2 Comments

Dude1234 Over a year ago

Thanks, however there appears to be a dip in the lines. Why does that happen?

Yohst Over a year ago

Thats the nature of smoothing/splines. Notices that smoothing data in this case really means making up data points in between the actual data points. That data does not exist/is fake. The fake data purely exists to provide a smooth plot and one must use some algorithm to pick that fake data to create 'smooth curves'. I used the bspline from scipy. Better is to take more data points naturally.

Collectives™ on Stack Overflow

Plotting multiple curve line plots in python with Matplotlib

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related