Streamlit_Bk7_Ch11_02.py

import numpy as np
import matplotlib.pyplot as plt
from matplotlib import cm
from sklearn import svm
import streamlit as st

p = plt.rcParams
p["font.sans-serif"] = ["Roboto"]
p["font.weight"] = "light"
p["ytick.minor.visible"] = True
p["xtick.minor.visible"] = True
p["axes.grid"] = True
p["grid.color"] = "0.5"
p["grid.linewidth"] = 0.5


with st.sidebar:
    st.title('线性核SVM')
    penalty = st.slider('Penalty', 0.0, 1.0, 0.1, 0.01)



# Create 20 points 
X = np.array([[0,2],[2,0],[4.5,0],[4.5,3],[3,4.5],
              [2,1],[1.5,3],[4,4],[0,3.5],[3,2.5],
              [-1,-1],[-3,-4.5],[-2,-3],[-4,-4],[0,-2.5],
              [-1,-3.5],[-3.5,-1],[0,-3.5],[-3,-1],[-4,0]])
y = [1] * 10 + [-1] * 10


clf = svm.SVC(kernel='linear', C=penalty)

# load data and fit the model
clf.fit(X, y)

# extract parameters of the decision boundary
w = clf.coef_[0]
w1 = w[0]
w2 = w[1]

b  = clf.intercept_[0]
x1_vec = np.linspace(-5, 5)

# decision boundary
x2_decision = -w1/w2*x1_vec - b/w2

# upper boundary of margin
x2_up   = -w1/w2*x1_vec - (b - 1)/w2

# lower boundary of margin
x2_down = -w1/w2*x1_vec - (b + 1)/w2

# visualization

fig, ax = plt.subplots(figsize = (5,5))
plt.plot(x1_vec, x2_decision, color=np.array([0, 68, 138])/255.)
plt.plot(x1_vec, x2_down, 'k--')
plt.plot(x1_vec, x2_up, 'k--')

plt.scatter(clf.support_vectors_[:, 0], clf.support_vectors_[:, 1], s=50,
            facecolors='none', edgecolors='k')

plt.scatter(X[:, 0], X[:, 1], c=y, cmap=cm.get_cmap('RdBu'),
            edgecolors='w')

x_1_2_min = -5
x_1_2_max = 5

YY, XX = np.meshgrid(x1_vec,x1_vec)
xy = np.vstack([XX.ravel(), YY.ravel()]).T
Z = clf.decision_function(xy).reshape(XX.shape)

cnt = plt.contourf(XX, YY, Z, levels = 20, cmap=cm.get_cmap('RdBu'),
                   alpha=0.5, linestyles=None)
for c in cnt.collections:
    c.set_edgecolor("face")
    
plt.axis('scaled')

ax.set_xlim(x_1_2_min, x_1_2_max)
ax.set_ylim(x_1_2_min, x_1_2_max)
ax.set_xlabel('$x_1$')
ax.set_ylabel('$x_2$')
plt.xticks(np.arange(x_1_2_min+1, x_1_2_max+1, 2))
plt.yticks(np.arange(x_1_2_min+1, x_1_2_max+1, 2))
ax.grid(linestyle='--', linewidth=0.25, color=[0.5,0.5,0.5])
plt.title("Penalty C = {:.2f}".format(penalty))

st.pyplot(fig)