#Import toy data and target
X = pd.read_csv('../../data/simulated/mvnsim/mvnsim' + dataset + '.csv',
sep=',',
header=0,
index_col=0).as_matrix()
y = np.load('../../data/simulated/mvnsim/target' + dataset + '.npy')
print(y)
print(y.shape)
print(X.shape)
#Plot initial data
plot_scatter(
X,
y,
'Initial data',
x_label='x coordinate',
y_label='y coordinate',
#output='save',
output='show',
#path='../../figs/out/%s/%s/initial.png' % (scriptname, dataset)
)
## PREPROCESSING ##
#Scale initial data to centre data
X_scaled = scale(X)
plot_scatter(
X_scaled,
y,
'Scaled data',
kpca_kernels = []
for kernel, abbreviation, kpca in kpcas:
if kernel == 'Laplacian KPCA':
X_kpca = kpca.fit_transform(K_lap)
else:
X_kpca = kpca.fit_transform(X_scaled)
p2.plot_scatter(
X_kpca,
y,
'First 2 principal components after %s' % kernel,
gamma=gamma,
x_label='Principal component 1',
y_label='Principal component 2',
#output = 'show',
#output='save',
#path='%s%s_%spca_gamma%s.png' % (filepath, nowtime, abbreviation, gamma)
)
#print('\nScatter plot of first two principal components after %sPCA for dataset %s saved.' % (kernel, dataset))
kpca_kernels.append(kernel)
# Declare names of models deployed and ROC AUC for each model
mdl_names = []
mean_aucs = []
for model_name, model_abv, model in models:
cv = StratifiedKFold(n_splits=10, random_state=10)
# Declare KPCA kernels deployed
kpca_kernels = []
for kernel, abbreviation, kpca in kpcas:
X_kpca = kpca.fit_transform(X_scaled)
plot_scatter(
X_kpca,
y,
'First 2 principal components after %sPCA' % kernel,
gamma=gamma,
x_label='Principal component 1',
y_label='Principal component 2',
#output = 'show',
output='save',
path='../../figs/out/%s/%s/%spca_gamma%s.png' %
(scriptname, dataset, abbreviation, gamma))
print(
'\nScatter plot of first two principal components after %sPCA for dataset %s saved.'
% (kernel, dataset))
X_kpca = kpca.fit_transform(X)
kpca_kernels.append(kernel)
# Declare names of models deployed
mdl_names = []
# Name of script to trace where images came from
scriptname = 'gs_tune_1_2'
#Select current toy dataset
dataset = '013'
#Import toy data and target
X = pd.read_csv('../../data/simulated/mvnsim/mvnsim' + dataset + '.csv', sep=',', header=0, index_col=0).as_matrix()
y = np.load('../../data/simulated/mvnsim/target' + dataset + '.npy')
#Plot initial data
plot_scatter(X,
y,
'Initial data',
x_label='x coordinate',
y_label='y coordinate',
output='save',
path='../../figs/out/%s/%s/initial.png' % (scriptname, dataset)
)
## PREPROCESSING ##
#Scale initial data to centre data
X_scaled = scale(X)
plot_scatter(X_scaled,
y, 'Scaled data',
x_label='x coordinate',
y_label='y coordinate',
#Create directory if directory does not exist
filepath = '../../figs/out/%s/%s/' % (scriptname, dataset)
if not os.path.exists(filepath):
os.makedirs(filepath)
#Import toy data and target
X = pd.read_csv('../../data/simulated/mvnsim/mvnsim' + dataset + '.csv', sep=',', header=0, index_col=0).as_matrix()
y = np.load('../../data/simulated/mvnsim/target' + dataset + '.npy')
#Plot initial data
plot_scatter(X,
y,
'Initial data',
x_label='x coordinate',
y_label='y coordinate',
#output='save',
#path='%sinitial.png' % filepath
output = 'show',
)
## PREPROCESSING ##
#Scale initial data to centre data
X_scaled = scale(X)
plot_scatter(X_scaled,
y, 'Scaled data',
x_label='x coordinate',
# To utilise precomputed kernel(s)
if kernel == 'Laplacian KPCA':
X_kpca = kpca.fit_transform(kpca_lap)
#elif kernel == 'Chi Squared KPCA':
# X_kpca = kpca.fit_transform(kpca_chi)
else:
X_kpca = kpca.fit_transform(X_scaled)
p2f.plot_scatter(
X_kpca,
y,
'First 2 principal components after %s' % kernel,
gamma=gamma,
x_label='Principal component 1',
y_label='Principal component 2',
#output = 'show',
output='save',
path='%s%s_%s_gamma%s.png' %
(filepath, nowtime, abbreviation, gamma),
writepath='%s%s_%s_%s_plottingdata.txt' %
(plotpath, nowtime, scriptname, dataset),
dataset=dataset,
kernel=kernel,
)
print(
'\nScatter plot of first two principal components after %s for dataset %s saved.'
% (kernel, dataset))
kpca_kernels.append(kernel)
# Declare names of models deployed
X, y = make_circles(n_samples=1000, factor=.3, noise=.05, random_state=12)
gamma = 5
x_kerns = [("linear", linear_kernel(X)),
("gaussian", rbf_kernel(X, gamma=gamma)),
("laplacian", laplacian_kernel(X, gamma=gamma)),
("cosine", cosine_similarity(X)), ("sigmoid", sigmoid_kernel(X))]
fig = plt.figure(figsize=(8, 6))
p2f.plot_scatter(X,
y,
'',
gamma=gamma,
x_label='x coordinate',
y_label='y coordinate',
cat1='Category 1',
cat0='Category 0',
output='save',
path='%sinit_scatter.png' % imgpath,
jspath='%sinit_scatter.js' % jspath)
for k_lab, k_X in x_kerns:
p2f.plot_scatter(k_X,
y,
'',
gamma=gamma,
x_label='x coordinate',
y_label='y coordinate',
cat1='Category 1',
