def visualize_zs(zs, labels, perplexity=30, learning_rate=200, n_iter=1000):
plt.clf()
plt.figure(figsize=(7,6))
points = TSNE(n_components=2, random_state=0, perplexity=perplexity, learning_rate=learning_rate, n_iter=n_iter).fit_transform(zs)
plt.scatter(points.transpose()[0], points.transpose()[1], s=20, c=labels)
plt.colorbar()
plt.show()
def visualize_zsp(zs, perplexities, labels=None, learning_rate=200, n_iter=2000):
plt.clf()
fig = plt.figure(figsize=(len(perplexities) * 4, 4))
for i, perplexity in enumerate(perplexities):
points = TSNE(n_components=2, random_state=0, perplexity=perplexity, learning_rate=learning_rate, n_iter=n_iter).fit_transform(zs)
ax = fig.add_subplot(1, len(perplexities), i + 1)
ax.scatter(points.transpose()[0], points.transpose()[1], s=5) if labels is None else ax.scatter(points.transpose()[0], points.transpose()[1], s=5, c=labels)
fig.show()
def CalculateTNSE(tdata):
u_tnse = TSNE(n_components=2).fit_transform(tdata.T)
u_tnse = u_tnse.transpose()
u_tnse = u_tnse[:2]
dataTNSE = np.concatenate((u_tnse[0][:,None],u_tnse[1][:,None]), axis = 1)
W_TNSE = tdata.dot(dataTNSE)
return W_TNSE
def compute_tsne__csv(input_filename, output_filename):
df = pd.read_csv(input_filename, header=0)
features_df = df[df.columns[2:12]]
X = np.array(features_df)
X_embedded = TSNE(
n_components=2,
early_exaggeration=48, # default is 12
learning_rate=100, # default is 200
verbose=0, # default is 0
).fit_transform(X)
X_embedded__transposed = X_embedded.transpose()
df.insert(14, 'tsne_0', X_embedded__transposed[0], True)
df.insert(15, 'tsne_1', X_embedded__transposed[1], True)
df.to_csv(output_filename, index=False)
df.to_csv(output_filename, index=False)
if __name__ == '__main__':
df = pd.read_csv(
'transitional_data/rnaseq_data__genes_x_timepoints__normalized__go_labeled.csv'
)
features_df = df[df.columns[2:12]]
X = np.array(features_df)
X_embedded = TSNE(
n_components=2,
early_exaggeration=48, # default is 12
learning_rate=100, # default is 200
verbose=2, # default is 0
).fit_transform(X)
X_embedded__transposed = X_embedded.transpose()
scatter_plot(X_embedded__transposed[0], X_embedded__transposed[1])
df.insert(14, 'tsne_0', X_embedded__transposed[0], True)
df.insert(15, 'tsne_1', X_embedded__transposed[1], True)
df.to_csv(
'transitional_data/rnaseq_data__genes_x_timepoints__normalized__go_labeled__tsne.csv',
index=False)
testVec = data[:, border:] # test data vectors
# initialize kohonen network
kohNet = nf.netInit(7)
# teach network on learnVec
kohNetLearned, clustLearnVec = nf.netLearn(kohNet, learnVec)
# test network
clustTestVec = nf.netClust(kohNetLearned, testVec)
# reduce data dimension to 3 by using TSNE (for possible visualising)
data_embedded = TSNE(n_components=3,
method='barnes_hut').fit_transform(data.transpose())
data_embedded = data_embedded.transpose()
# output results
clustVec = np.concatenate((clustLearnVec, clustTestVec))
print("\nClusters of data\n", clustVec, "\n")
print("Number of classes: ", kohNetLearned['numClusters'])
print("\nEmbedded data\n", data_embedded)
print('\n-- The algorithm took', np.float16(time.time() - start_time),
'seconds to complete --') # execution time
# In[ ]:
import numpy as np import matplotlib.pyplot as plt from sklearn.manifold import TSNE X = np.array([[0, 1, 1], [1, 0, 1], [1, 1, 0]]) X_embedded = TSNE(n_components=2).fit_transform(X) Y = X_embedded.transpose() plt.plot(Y[0], Y[1], "p") plt.show()