def reduce_dims(X=None,
y=None,
data_path=None,
algorithm='pca',
perplexity=50,
labels=['M', 'R', 'HC', 'V', 'PO']):
"""visualize the data in a latent space with reduced dimensions"""
if data_path is not None:
X, y = load_tma_data(data_path)
X = X.reshape(X.shape[0], X.shape[1] * X.shape[2])
print("Number of Training Samples : ", X.shape[0])
# standardize the data
sc = StandardScaler()
X = sc.fit_transform(X)
# reduce the dimensionality of the data
if algorithm == 'pca':
pca = PCA(n_components=2)
X_reduced = pca.fit_transform(X)
if algorithm == 'tsne':
tsne = TSNE(n_components=2,
perplexity=perplexity,
random_state=0,
verbose=True)
X_reduced = tsne.fit_transform(X)
# plot the latent space
plot_latent_space(X_reduced, y, labels)
X_test = np.vstack((X_test, X1i[21 * 7:]))
X_test = np.vstack((X_test, X2i[21 * 7:]))
y_train = np.concatenate((y_train, i * np.ones((21 * 14, ))))
y_test = np.concatenate((y_test, i * np.ones((21 * 6, ))))
X1, y1 = X_train, y_train
X2, y2 = X_test, y_test
if visualizeLatentSpace:
X = np.vstack((X1, X2))
y = np.hstack((y1, y2))
print(X.shape)
print(y.shape)
Xr = TSNE(n_components=2,
perplexity=50).fit_transform(X.reshape(X.shape[0], 8 * 8))
plot_latent_space(Xr, y, labels=class_labels)
## shuffle
X1, y1 = shuffle(X1, y1, random_state=0)
X2, y2 = shuffle(X2, y2, random_state=0)
## create graph dataset
Xg1 = createGraphDataset(X1, y1)
Xg2 = createGraphDataset(X2, y2)
## define model
model = Net()
## define the optimizer
optimizer = torch.optim.Adam(model.parameters(), lr=lr)
GL = DiffusionGraphLearn(
X,
p=p,
beta_1=beta_1,
beta_2=beta_2,
verbosity=True
)
W = GL.findGraphLaplacian()
W /= np.max(W)
# W[W<0.4] = 0
Ws = np.vstack((Ws, np.expand_dims(W, 0)))
y.append(i)
Xr = PCA(n_components=2).fit_transform(Ws.reshape(Ws.shape[0], 8*8))
plot_latent_space(Xr, y, labels=labels)
if combinedReduceDims:
df = pd.read_csv(onsets_file, delimiter=',')
Ws = np.empty((0, 8, 8))
y = []
labels = ["M", "R", "HC", "V", "PO"]
for i in range(len(labels)):
file1 = os.path.join(data_path, "{}_2.csv".format(labels[i]))
file2 = os.path.join(data_path, "{}_3.csv".format(labels[i]))
data1 = read_from_csv(id, file1)
data1 = preprocess(data1, fs, lpfreq=1)
data2 = read_from_csv(id, file2)
data2 = preprocess(data2, fs, lpfreq=1)
onsets1 = df[['{}_2'.format(labels[i])]].values.squeeze()
onsets2 = df[['{}_3'.format(labels[i])]].values.squeeze()