norm 연습

L1, L2 규제를 공부하다가 연습으로 norm을 시각화해봄

코드

import matplotlib.pyplot as plt
import numpy as np
import math
x1 = np.arange(-10, 10, 0.1)
x2 = np.arange(-10, 10, 0.1)
xx1,xx2 = np.meshgrid(x1, x2)
pp = [0.5, 1, 2, 3, 4, 5]
fig, ax = plt.subplots(nrows=1, ncols=len(pp), figsize=(15,3))
for idx, p in enumerate(pp) :
    Z = np.array([(abs(a)**p + abs(b)**p)**p for a, b in zip(xx1.ravel(), xx2.ravel())])
    ax[idx].contourf(xx1, xx2, Z.reshape(xx1.shape), alpha=0.3)
    ax[idx].set_xlim(xx1.min(), xx1.max())   
    ax[idx].set_title("p = {}".format(p))
    
ax[0].set_ylim(xx2.min(), xx2.max())
plt.show()

결과

참고

https://ekamperi.github.io/machine%20learning/2019/10/19/norms-in-machine-learning.html