def test_plot20(self):
np.random.seed(seed)
Xa = np.array([[i, 0.0] for i in np.linspace(-10.0, 10.0, 40)])
Xb = np.array([[i, 0.2] for i in np.linspace(-10.0, 10.0, 20)])
Xc = np.array([[i, -0.2] for i in np.linspace(-10.0, 10.0, 20)])
ya = ['A' for i in range(40)]
yb = ['B' for i in range(20)]
yc = ['C' for i in range(20)]
X = np.concatenate([Xa, Xb, Xc], axis=0)
y = np.concatenate([ya, yb, yc])
nca = NCA()
# rcParams['text.usetex'] = 'True' # Latex installed needed
# rcParams['text.latex.preamble'] = r'\usepackage{amsmath}'
nca.fit(X, y)
dml_multiplot(
X,
y,
nrow=1,
ncol=3,
ks=[1, 1, 1],
fitted=True,
dmls=[None, nca, nca],
transforms=[False, False, True],
cmap="gist_rainbow",
# subtitles=[r'$M=\begin{pmatrix}1 & 0 \\ 0 & 1 \end{pmatrix}$',
# r'$M \approx \begin{pmatrix} 0 & -0.004 \\ -0.004 & 27.5 \end{pmatrix}$',
# r'$L \approx \begin{pmatrix} -0.0001 & 0.073 \\ -0.0008 & 5.24 \end{pmatrix}$'],
figsize=(18, 6))
self.newsave()
plt.close()
def test_plot22(self):
np.random.seed(seed)
X, y = toy_datasets.circular_toy_dataset(rads=[1, 2],
samples=[200, 200],
noise=[0.0, 0.0])
y = y.astype(str)
yy = y.astype(str)
yy[40:200] = '?'
yy[220:] = '?'
knn1 = KNeighborsClassifier(1)
knn1.fit(X[np.isin(yy, ['0', '1'])], yy[np.isin(yy, ['0', '1'])])
knn2 = KNeighborsClassifier(1)
knn2.fit(X, y)
dml_multiplot(X,
yy,
clfs=[knn1, knn2],
label_colors=['red', 'blue', 'lightgray'],
fitted=True,
figsize=(12, 6))
self.newsave()
plt.close()
def test_plot23(self):
np.random.seed(seed)
X, y = iris_data()
y = [
'Setosa' if yi == 0 else 'Versicolor' if yi == 1 else 'Virginica'
for yi in y
]
X = pd.DataFrame(X,
columns=[
'Sepal Length', 'Sepal Width', 'Petal Length',
'Petal Width'
])
arr = ['Sepal Length', 'Sepal Width']
X = X[arr]
dml_multiplot(X,
y,
ks=[1, 30],
cmap='rainbow',
subtitles=[r'$k=1$', r'$k=30$'],
figsize=(12, 6))
self.newsave()
plt.close()
def test_plot16(self):
np.random.seed(seed)
X, y = toy_datasets.balls_toy_dataset(centers=[[-1.0, 0.0], [0.0, 0.0],
[1.0, 0.0]],
rads=[0.3, 0.3, 0.3],
samples=[50, 50, 50],
noise=[0.1, 0.1, 0.1])
y[y == 2] = 0
y = y.astype(int)
ncm = NearestCentroid()
ncmc = NCMC_Classifier(centroids_num=[2, 1])
dml_multiplot(X,
y,
nrow=1,
ncol=2,
clfs=[ncm, ncmc],
cmap='rainbow',
subtitles=['NCM', 'NCMC'],
figsize=(6, 3))
self.newsave()
plt.close()
def test_plot9(self):
np.random.seed(seed)
X, y = digits_data()
selected = np.where(np.isin(y, [0, 1, 3, 4, 6, 9]))[0]
X, y = X[selected, :], y[selected]
lda = LDA(num_dims=2)
anmm = ANMM(num_dims=2, n_friends=1, n_enemies=1)
dml_multiplot(
X,
y,
nrow=2,
ncol=2,
ks=[1, 1, 11, 11],
dmls=[lda, anmm, lda, anmm],
title="Comparing DMLS",
subtitles=["k=1, LDA", "k=1, ANMM", "k=11, LDA", "k=11, ANMM"],
cmap="gist_rainbow",
plot_points=True,
figsize=(20, 16))
self.newsave()
plt.close()
def test_plot12(self):
np.random.seed(seed)
X = np.array([[0.0, 0.1], [0.5, 0.1], [-0.5, -0.1], [1.0, 0.2],
[-1.0, -0.1], [0.1, 1.0], [-0.1, 1.0], [0.1, -1.0],
[-0.1, -1.0]])
y = np.array([0, 0, 0, 0, 0, 1, 1, 2, 2])
lmnn = LMNN(max_iter=1000,
learning_rate="adaptive",
eta0=1.0,
k=1,
mu=0.5,
tol=1e-15,
prec=1e-15)
dml_multiplot(X,
y,
2,
2,
ks=[1, 1, 1],
dmls=[None, lmnn, lmnn],
transforms=[True, True, False],
title="LMNN",
subtitles=["Original", "Transformed", "Region LMNN+KNN"],
cmap="rainbow",
figsize=(20, 20))
self.newsave()
knn_plot(X, y, k=1, cmap="gist_rainbow", figsize=(15, 8))
self.newsave()
knn_plot(X,
y,
k=1,
cmap="gist_rainbow",
figsize=(15, 8),
transformer=np.array([[0.0, 0.0], [0.0, 3.0]]),
transform=False)
self.newsave()
plt.close()
def test_plot18(self):
np.random.seed(seed)
svm = SVC(kernel='linear')
Xa = np.array([[1.8 * np.random.random() + -0.9, 0.0]
for i in range(40)])
Xb = np.array([[0.4 * np.random.random() + 1.1, 0.0]
for i in range(20)])
Xc = np.array([[0.4 * np.random.random() - 1.5, 0.0]
for i in range(20)])
X = np.concatenate([Xa, Xb, Xc], axis=0)
y = np.empty(X.shape[0])
y[np.abs(X[:, 0]) > 1] = 1
y[np.abs(X[:, 0]) < 1] = -1
y = y.astype(int)
f, ax = plt.subplots(sharex='row', sharey='row', figsize=(6, 3))
sc = ax.scatter(X[:, 0],
X[:, 1],
c=y,
cmap='rainbow',
s=20,
edgecolor='k')
handles = [
plt.plot([],
color=sc.get_cmap()(sc.norm(c)),
ls="",
marker="o",
mec="k")[0] for c in [-1, 1]
]
ax.legend(handles, [-1, 1], loc="lower right")
self.newsave()
f1, ax1 = plt.subplots(figsize=(4, 4))
ax1.set_xlim([-1.6, 1.6])
ax1.set_ylim([-2.0, 2.0])
sc = ax1.scatter(X[:, 0],
X[:, 1],
c=y,
cmap='rainbow',
s=20,
edgecolor='k')
handles = [
plt.plot([],
color=sc.get_cmap()(sc.norm(c)),
ls="",
marker="o",
mec="k")[0] for c in [-1, 1]
]
ax1.legend(handles, [-1, 1], loc="lower right")
X[:, 1] = X[:, 0] * X[:, 0]
L = np.array([[1, 0], [0, 0]])
svq = SVC(kernel='poly', degree=2)
dml_multiplot(X,
y,
nrow=1,
ncol=2,
clfs=[svm, svq],
transformers=[None, L],
transforms=[False, True],
cmap='rainbow',
figsize=(12, 6))
self.newsave()
f2, ax2 = plt.subplots(figsize=(4, 4))
ax2.set_xlim([-1.6, 1.6])
ax2.set_ylim([-0.1, 2.5])
ax2.scatter(X[:, 0], X[:, 1], c=y, cmap='rainbow', s=20, edgecolor='k')
handles = [
plt.plot([],
color=sc.get_cmap()(sc.norm(c)),
ls="",
marker="o",
mec="k")[0] for c in [-1, 1]
]
ax2.legend(handles, [-1, 1], loc="lower right")
self.newsave()
dml_plot(X,
y,
clf=svm,
transform=False,
cmap='rainbow',
figsize=(4, 4),
xrange=[-1.6, 1.6],
yrange=[-0.1, 2.5])
self.newsave()
dml_plot(X,
y,
clf=svq,
transformer=L,
transform=True,
cmap='rainbow',
xrange=[-1.6, 1.6],
yrange=[-2.0, 2.0],
figsize=(4, 4))
self.newsave()
plt.close()