In [1]:
# Code attribution: Yiyin Shen, Tyler Caraza-Harter
# Imports
import sklearn
import sklearn.datasets
import sklearn.svm
import sklearn.neural_network
import sklearn.model_selection
import sklearn.inspection
import seaborn
import numpy
import matplotlib.pyplot as plt
import warnings
warnings.filterwarnings('ignore')
In [2]:
# Create a random dataset with two circles
x, y = sklearn.datasets.make_circles(n_samples = 100, noise = 0.25, factor = 0.25, random_state = 42)
seaborn.relplot(x = x[:, 0], y = x[:, 1], hue = y)
Out[2]:
<seaborn.axisgrid.FacetGrid at 0x2551e39a310>
In [3]:
# Neural Network
nn = sklearn.neural_network.MLPClassifier(hidden_layer_sizes = [20, 20])
nn.fit(x, y)
plot = sklearn.inspection.DecisionBoundaryDisplay.from_estimator(nn, x, response_method = "predict_proba")
plot.ax_.scatter(x[:, 0], x[:, 1], c = y, edgecolor = "black")
Out[3]:
<matplotlib.collections.PathCollection at 0x255204b08d0>
In [4]:
# Kernel SVM
svm = sklearn.svm.SVC(kernel = "rbf")
svm.fit(x, y)
plot = sklearn.inspection.DecisionBoundaryDisplay.from_estimator(svm, x, response_method = "predict")
plot.ax_.scatter(x[:, 0], x[:, 1], c = y, edgecolor = "black")
Out[4]:
<matplotlib.collections.PathCollection at 0x255204f1a10>
In [5]:
# Directly compute accuracy (not the correct way)
y_pred_nn = nn.predict(x)
y_pred_svm = svm.predict(x)
print([sum(y_pred_nn == y), sum(y_pred_svm == y)])
[90, 94]
In [6]:
# Cross validation accuracy
acc_nn = sklearn.model_selection.cross_val_score(nn, x, y, cv = 5)
acc_svm = sklearn.model_selection.cross_val_score(svm, x, y, cv = 5)
print([round(numpy.average(acc_nn * 100)), round(numpy.average(acc_svm * 100))])
[87, 89]
In [7]:
# Cross validation distribution (mean and variance)
fig, ax = plt.subplots()
ax.bar([0, 2], [acc_nn.mean(), acc_svm.mean()])
ax.errorbar([0, 2], [acc_nn.mean(), acc_svm.mean()], yerr = [acc_nn.std(), acc_svm.std()], elinewidth = 5, color = "red", linestyle='none')
Out[7]:
<ErrorbarContainer object of 3 artists>
