Huber Loss function

Huber loss是为了增强平方误差损失函数（squared loss function）对噪声（或叫离群点，outliers）的鲁棒性提出的。

Definition

Lδ(a)={12a2,δ⋅(|a|−12δ),for |a|≤δ,otherwise.Lδ(a)={12a2,for |a|≤δ,δ⋅(|a|−12δ),otherwise.

参数aa通常表示residuals，也即(y−y^)(y−y^)或者写作(y−f(x))(y−f(x))，当a=y−f(x)a=y−f(x)时，上述形式可以拓展为： 

Lδ(y,f(x))={12(y−f(x))2,δ⋅(|y−f(x)|−12δ),for |y−f(x)|≤δotherwise.Lδ(y,f(x))={12(y−f(x))2,for |y−f(x)|≤δδ⋅(|y−f(x)|−12δ),otherwise.

visualization

import numpy as np
import matplotlib.pyplot as plt

def huber_loss(e, d):
    return (abs(e)<=d)*e**2/2 + (e>d)*d*(abs(e)-d/2)

plt.figure(figsize=(6, 4.5), facecolor='w', edgecolor='k')
x = np.arange(-20, 20)
plt.plot(x, x**2/2, label='squared loss', lw=2, 'g')
for d in (10, 5, 3, 1):
    plt.plot(x, huber_loss(x, d), label=r'huber loss: $\delta$={}'.format(d), lw=2)

plt.legend(loc='best', frameon=False)
plt.xlabel('standard deviation')
plt.ylabel('loss')
plt.show()

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14
• 15
• 16

 

References

[1] Huber Loss Wikipedia