def test_kruskal_stress():
actual_data = chooseData('scurve')
fitted_data = chooseEmbedder('isomap', actual_data)
fitted_data.embed(use_cache=True)
evaluator = StressEvaluator(actual_data.df, 100)
assert evaluator.kruskal(fitted_data.em) > 0
def test_global_ranking_stress():
actual_data = chooseData('scurve')
fitted_data = chooseEmbedder('isomap', actual_data)
fitted_data.embed(use_cache=True)
evaluator = StressEvaluator(actual_data.df, 100)
assert evaluator.middle_ranking(fitted_data.em, 15, is_z_value=False) > 0
def test_kernel_pca_clustered_swissroll():
logging.info("start kernel_pca test using clustered-swissroll data")
logging.info("start generating data")
kernel_pca = KernelPCAEmbedder(chooseData("clustered_swissroll"))
logging.info("start embedding")
kernel_pca.embed(use_cache=True)
logging.info("finish all process")
assert kernel_pca.em.shape == (1000, 2)
def test_kernel_pca_basic_cluster():
logging.info("start kernel_pca test using basic_cluster data")
logging.info("start generating data")
kernel_pca = KernelPCAEmbedder(chooseData("basic_cluster"))
logging.info("start embedding")
kernel_pca.embed(use_cache=True)
logging.info("finish all process")
assert kernel_pca.em.shape == (801, 2)
def test_isomap_and_nmf():
isomap = IsomapEmbedder(chooseData("pokemon"))
isomap.embed(dim=2, use_cache=True)
assert isomap.em.shape == (801, 2)
nmf = NMFReducer(isomap.data, isomap)
assert nmf.df.shape == (801, 2)
nmf.reduce(dim=2, use_cache=True)
assert nmf.rd.shape == (801, 2)
def getFigure(data_key, embedder_key, reducer_key):
sc_data = chooseData(data_key)
embedder = chooseEmbedder(embedder_key, sc_data)
embedder.embed()
reducer = chooseReducer(reducer_key, sc_data, embedder)
reducer.reduce()
return px.scatter(reducer.rd,
x='col0',
y='col1',
color=sc_data.color,
title=f'{data_key} | {reducer.class_key}')
def get_embedding(data_key: str, embedder_key: str) -> DataFrame:
"""Get embedding
Args:
data_key (str): Name of dataset.
embedder_key (str): Name of embedder.
Returns:
DataFrame: Embedding results.
"""
sc_data = chooseData(data_key)
embedder = chooseEmbedder(embedder_key, sc_data)
embedder.embed()
return embedder.em
def test_n_mds():
n_mds = N_MDSEmbedder(chooseData("pokemon"))
n_mds.embed(use_cache=True)
assert n_mds.em.shape == (801, 2)
def test_t_sne_basic_cluster():
tsne = TSNEEmbedder(chooseData("basic_cluster"))
tsne.embed(use_cache=True, dim=2)
assert tsne.em.shape == (801, 2)
def test_t_sne_pokemon():
tsne = TSNEEmbedder(chooseData("pokemon"))
tsne.embed(use_cache=True, dim=2)
assert tsne.em.shape == (801, 2)
def test_nmf():
nmf = NMFReducer(chooseData("pokemon"))
nmf.reduce(use_cache=True)
assert nmf.rd.shape == (801, 2)
def test_locally_linear():
locally_linear = LocallyLinearEmbedder(chooseData("pokemon"))
locally_linear.embed(use_cache=True)
assert locally_linear.em.shape == (801, 2)
def test_kernel_pca():
kernel_pca = KernelPCAEmbedder(chooseData("pokemon"))
kernel_pca.embed(use_cache=True)
assert kernel_pca.em.shape == (801, 2)
margin=dict(r=20, l=10, b=10, t=10),
)
fig3d.show()
# Create 2d figure After Dinmensionality Reduction
fig2d = visualize_2d(reducer.rd)
fig2d.update_layout(title_text="After Diminsionality Reduction")
fig2d.show()
if __name__ == "__main__":
# Get 3d animations
# Run these in Jupyter or something
which_data = "basic_cluster"
embedder = chooseEmbedder("t_sne", (chooseData(which_data)))
embedder.embed(dim=3, use_cache=True)
reducer = chooseReducer("pca", chooseData(which_data), embedder)
reducer.reduce(dim=2, save_rd=False)
# visualize_3d_to_2d_projection(embedder, reducer)
# Double filter
query1 = "col1<0 & col2>0 & col2>=0.5"
query0 = "*"
# For Dimensionality Reduction to 2D
reducer.setRds(query0=query0, query1=query1)
print(reducer.getRdsDf())
fig2d = px.scatter(
reducer.getRdsDf(),
def test_isomap():
isomap = IsomapEmbedder(chooseData("pokemon"))
isomap.embed(use_cache=True)
assert isomap.em.shape == (801, 2)
def test_l_em():
l_em = Laplacian_EigenmapsEmbedder(chooseData("pokemon"))
l_em.embed(use_cache=True)
assert l_em.em.shape == (801, 2)
def test_pca():
pca = PCAReducer(chooseData("pokemon"))
pca.reduce(use_cache=True)
assert pca.rd.shape == (801, 2)
