def tsne(df_anim, ndim):
""" Takes a df of animation, reduces the 17d to ndim dimensions """
joints_values = df_anim.loc[:, joints_names].values
X = np.array(joints_values)
X_embedded = TSNE(n_components=ndim,
perplexity=100.0,
learning_rate=30,
n_iter=50000,
init='random',
random_state=4).fit_transform(X)
return pd.Series(X_embedded.reshape([
100,
]))
def cluster(features, num_objs, n_clusters, save_path):
name = 'features'
centers = {}
for label in range(num_objs):
feat = features[label]
if feat.shape[0]:
n_feat_clusters = min(feat.shape[0], n_clusters)
if n_feat_clusters < n_clusters:
print(label)
kmeans = KMeans(n_clusters=n_feat_clusters, random_state=0).fit(feat)
if n_feat_clusters == 1:
centers[label] = kmeans.cluster_centers_
else:
one_dimension_centers = TSNE(n_components=1).fit_transform(kmeans.cluster_centers_)
args = np.argsort(one_dimension_centers.reshape(-1))
centers[label] = kmeans.cluster_centers_[args]
save_name = os.path.join(save_path, name + '_clustered_%03d.npy' % n_clusters)
np.save(save_name, centers)
print('saving to %s' % save_name)
S = np.array([None]*df.shape[0])
for i in range(0, len(df)):
S[i] = df.samples[i][int(0.5 + df.cfd_trig_rise[i]/1000)-20: int(0.5 + df.cfd_trig_rise[i]/1000)+waveFormLegth-20].astype(np.float64)
return S
St = get_samples(df_test)
x_test=np.stack(St)#df_test.samples)
x_test=x_test.reshape(len(x_test), window_width, 1)
model_path='weights-improvement-01-0.79.hdf5'#model_zero.hdf5'
model=keras.models.load_model('../CNN_models/%s'%model_path)
predictions = model.predict(x_test)
df_test['pred'] = predictions
int_layer_model = Model(inputs=model.input, outputs=model.get_layer('flat').output)
params = int_layer_model.predict(x_test)
params = StandardScaler().fit_transform(params)
X_embedded = TSNE(n_components=2, init='pca', n_iter=1000, verbose=2, perplexity=20, method='barnes_hut', learning_rate=10, early_exaggeration=2).fit_transform(params)
# pca = PCA(n_components=30)
# pca.fit(params)
# print(pca.explained_variance_ratio_)
# print(pca.explained_variance_ratio_.sum())
# params_new = pca.transform(params)
# params_new = params_new.reshape(params_new.shape[1], params_new.shape[0])
X1=X_embedded.reshape(X_embedded.shape[1], X_embedded.shape[0])[0]
X2=X_embedded.reshape(X_embedded.shape[1], X_embedded.shape[0])[1]
plt.scatter(X1, X2, c=df_test.pred, cmap='viridis')
plt.colorbar()
plt.show()
columns = batch_size // rows if batch_size // rows < 8 else 8
else:
rows = 2
columns = 2
span = np.linspace(-4, 4, rows)
grid = np.dstack(np.meshgrid(span, span)).reshape(-1, 2)
# If you want to use a dimensionality reduction method you can use
# for example PCA by projecting on two principal dimensions
""" PCA """
# z = PCA(n_components=2).fit_transform(z.reshape(-1,latent_features))
# z = z.reshape(batch_size,num_samples,2)
""" TSNE """
print(z.shape)
z = TSNE(n_components=2).fit_transform(z.reshape(-1, latent_features))
z = z.reshape(batch_size, num_samples**2, 2)
# z = z[:,1,] # We do only want to plot one z instead of |num_samples|
colors = iter(plt.get_cmap('Set1')(np.linspace(0, 1.0, len(classes))))
for c in classes:
ax.scatter(*z[c == y.numpy()].reshape(-1, 2).T,
c=next(colors),
marker='o',
label=c)
if show_sampling_points:
ax.scatter(*grid.T,
color="k",
marker="x",
alpha=0.5,
imzml = imzmlio.open_imzml(inputname)
spectra = imzmlio.get_full_spectra(imzml)
max_x = max(imzml.coordinates, key=lambda item: item[0])[0]
max_y = max(imzml.coordinates, key=lambda item: item[1])[1]
max_z = max(imzml.coordinates, key=lambda item: item[2])[2]
image = imzmlio.get_images_from_spectra(spectra, (max_x, max_y, max_z))
mzs, intensities = imzml.getspectrum(0)
else:
image = sitk.GetArrayFromImage(sitk.ReadImage(inputname)).T
mzs = [i for i in range(image.shape[2])]
mzs = np.asarray(mzs)
image = image[..., mzs >= threshold]
print(image.shape)
mzs = mzs[mzs >= threshold]
mzs = np.around(mzs, decimals=2)
mzs = mzs.astype(str)
shape = image.shape[:-1]
image_norm = imzmlio.normalize(image)
image_norm = fusion.flatten(image_norm, is_spectral=True).T
n = 3
X_tsne = TSNE(n_components=n, random_state=0).fit_transform(image_norm)
X_tsne = imzmlio.normalize(X_tsne)
X_tsne = X_tsne.reshape(shape + (3, ))
tsne_itk = sitk.GetImageFromArray(X_tsne, isVector=True)
sitk.WriteImage(tsne_itk, outputname)
