def plot_component_variance(x, y): prim_basis = PrimitiveBasis(n_states=3, domain=[0, 2]) model = MKSHomogenizationModel(basis=prim_basis) model.n_components = 20 model.fit(x, y, periodic_axes=[0, 1]) # Draw the plot containing the PCA variance accumulation draw_component_variance(model.dimension_reducer.explained_variance_ratio_)
y=np.ndarray(shape=(samples*end_of_transience, metadata[0]['x']*metadata[0]['y']*2))
y_ind = 0
for metadatum in metadata:
# Load data frames
print "--->Loading: " + metadatum['filename']
al_chunk = sp.load_data('data/test/'+metadatum['filename'])
corrs = sp.compute_correlations(al_chunk[0:end_of_transience, :, :],
correlations=[(0,0), (1,1)],
compute_flat=False)
for row in corrs:
y[y_ind] = row.flatten()
y_ind+=1
# Get some PCA components!
pca = PCA(n_components=12)
y_pca = pca.fit_transform(y)
draw_component_variance(pca.explained_variance_ratio_)
# Zip and pickle PCA comps
with gzip.GzipFile('data/pca_scores_transient.pgz', 'w') as f:
pickle.dump(y_pca, f)
sp.write_pca_to_csv(y_pca, '_transients')
time_y = np.ndarray((samples, end_of_transience))
row_ind = 0;
for i in range(0, y_pca.shape[0], end_of_transience):
time_y[row_ind, :] = y_pca[i:(i+end_of_transience), 0]
row_ind += 1
mse = leave_one_out(x, time_y)
print "MSE AVE: " + str(np.mean(mse))
def plot_component_variance(y): pca = PCA(n_components=12) y_pca = pca.fit_transform(y) # Draw the plot containing the PCA variance accumulation draw_component_variance(pca.explained_variance_ratio_)
samples = len(metadata)
x = np.ndarray(shape=(samples, 2))
y = np.ndarray(shape=(samples * end_of_transience,
metadata[0]['x'] * metadata[0]['y'] * 2))
y_ind = 0
x_ind = 0
for metadatum in metadata:
# Load data frames
print "--->Loading: " + metadatum['filename']
al_chunk = sp.load_data('data/test/' + metadatum['filename'])
corrs = sp.compute_correlations(
al_chunk[0:end_of_transience, :, :],
correlations=[(0, 0), (1, 1)],
compute_flat=False)
for row in corrs:
y[y_ind] = row.flatten()
y_ind += 1
x[x_ind, 0] = metadatum['ag']
x[x_ind, 1] = metadatum['sv']
x_ind += 1
# Get some PCA components!
pca = PCA(n_components=12)
y_pca = pca.fit_transform(y)
draw_component_variance(pca.explained_variance_ratio_)
# Zip and pickle PCA comps
with gzip.GzipFile('data/pca_scores_transient.pgz', 'w') as f:
pickle.dump(y_pca, f)
sp.write_pca_to_csv(y_pca, '_transients')
