def test_umap_data_formats(input_type, should_downcast, nrows, n_feats, name):
dtype = np.float32 if not should_downcast else np.float64
n_samples = nrows
n_feats = n_feats
if name == 'digits':
# use the digits dataset for unit test
digits = datasets.load_digits(n_class=9)
X = digits["data"].astype(dtype)
else:
X, y = datasets.make_blobs(n_samples=n_samples,
n_features=n_feats,
random_state=0)
umap = cuUMAP(n_neighbors=3, n_components=2, verbose=False)
if input_type == 'dataframe':
X_pd = pd.DataFrame({'fea%d' % i: X[0:, i] for i in range(X.shape[1])})
X_cudf = cudf.DataFrame.from_pandas(X_pd)
embeds = umap.fit_transform(X_cudf, convert_dtype=True)
assert type(embeds) == cudf.DataFrame
else:
embeds = umap.fit_transform(X)
assert type(embeds) == np.ndarray
def test_umap_data_formats(input_type, should_downcast, nrows, n_feats, name):
dtype = np.float32 if not should_downcast else np.float64
n_samples = nrows
n_feats = n_feats
if name == 'digits':
# use the digits dataset for unit test
digits = datasets.load_digits(n_class=9)
X = digits["data"].astype(dtype)
else:
X, y = datasets.make_blobs(n_samples=n_samples,
n_features=n_feats,
random_state=0)
umap = cuUMAP(n_neighbors=3, n_components=2, verbose=False)
embeds = umap.fit_transform(X)
assert type(embeds) == np.ndarray
if enable_gpu:
kmeans_float = cuml.KMeans(n_clusters=n_clusters)
else:
kmeans_float = sklearn.cluster.KMeans(n_clusters=n_clusters)
kmeans_float.fit(df_fingerprints)
print('Runtime Kmeans time (hh:mm:ss.ms) {}'.format(datetime.now() -
task_start_time))
# UMAP
task_start_time = datetime.now()
if enable_gpu:
umap = cuml.UMAP(n_neighbors=100, a=1.0, b=1.0, learning_rate=1.0)
else:
umap = umap.UMAP()
Xt = umap.fit_transform(df_fingerprints)
print('Runtime UMAP time (hh:mm:ss.ms) {}'.format(datetime.now() -
task_start_time))
if enable_gpu:
df_fingerprints.add_column('x', Xt[0].to_array())
df_fingerprints.add_column('y', Xt[1].to_array())
df_fingerprints.add_column('cluster', kmeans_float.labels_)
else:
df_fingerprints['x'] = Xt[:, 0]
df_fingerprints['y'] = Xt[:, 1]
df_fingerprints['cluster'] = kmeans_float.labels_
# start dash
v = chemvisualize.ChemVisualization(df_fingerprints.copy(),
n_clusters,
def visualize_features(classes, problem_type, curdir, default_features,
balance_data, test_size):
# make features into label encoder here
features, feature_labels, class_labels = get_features(
classes, problem_type, default_features, balance_data)
# now preprocess features for all the other plots
os.chdir(curdir)
le = preprocessing.LabelEncoder()
le.fit(class_labels)
tclass_labels = le.transform(class_labels)
# process features to help with clustering
se = preprocessing.StandardScaler()
t_features = se.fit_transform(features)
X_train, X_test, y_train, y_test = train_test_split(features,
tclass_labels,
test_size=test_size,
random_state=42)
# print(len(features))
# print(len(feature_labels))
# print(len(class_labels))
# print(class_labels)
# GET TRAINING DATA DURING MODELING PROCESS
##################################
# get filename
# csvfile=''
# print(classes)
# for i in range(len(classes)):
# csvfile=csvfile+classes[i]+'_'
# get training and testing data for later
# try:
# print('loading training files...')
# X_train=pd.read_csv(prev_dir(curdir)+'/models/'+csvfile+'train.csv')
# y_train=X_train['class_']
# X_train.drop(['class_'], axis=1)
# X_test=pd.read_csv(prev_dir(curdir)+'/models/'+csvfile+'test.csv')
# y_test=X_test['class_']
# X_test.drop(['class_'], axis=1)
# y_train=le.inverse_transform(y_train)
# y_test=le.inverse_transform(y_test)
# except:
# print('error loading in training files, making new test data')
# Visualize each class (quick plot)
##################################
visualization_dir = 'visualization_session'
try:
os.mkdir(visualization_dir)
os.chdir(visualization_dir)
except:
shutil.rmtree(visualization_dir)
os.mkdir(visualization_dir)
os.chdir(visualization_dir)
objects = tuple(set(class_labels))
y_pos = np.arange(len(objects))
performance = list()
for i in range(len(objects)):
performance.append(class_labels.count(objects[i]))
plt.bar(y_pos, performance, align='center', alpha=0.5)
plt.xticks(y_pos, objects)
plt.xticks(rotation=90)
plt.title('Counts per class')
plt.ylabel('Count')
plt.xlabel('Class')
plt.tight_layout()
plt.savefig('classes.png')
plt.close()
# set current directory
curdir = os.getcwd()
# ##################################
# # CLUSTERING!!!
# ##################################
##################################
# Manifold type options
##################################
'''
"lle"
Locally Linear Embedding (LLE) uses many local linear decompositions to preserve globally non-linear structures.
"ltsa"
LTSA LLE: local tangent space alignment is similar to LLE in that it uses locality to preserve neighborhood distances.
"hessian"
Hessian LLE an LLE regularization method that applies a hessian-based quadratic form at each neighborhood
"modified"
Modified LLE applies a regularization parameter to LLE.
"isomap"
Isomap seeks a lower dimensional embedding that maintains geometric distances between each instance.
"mds"
MDS: multi-dimensional scaling uses similarity to plot points that are near to each other close in the embedding.
"spectral"
Spectral Embedding a discrete approximation of the low dimensional manifold using a graph representation.
"tsne" (default)
t-SNE: converts the similarity of points into probabilities then uses those probabilities to create an embedding.
'''
os.mkdir('clustering')
os.chdir('clustering')
# tSNE
plt.figure()
viz = Manifold(manifold="tsne", classes=set(classes))
viz.fit_transform(np.array(features), tclass_labels)
viz.poof(outpath="tsne.png")
plt.close()
# os.system('open tsne.png')
# viz.show()
# PCA
plt.figure()
visualizer = PCADecomposition(scale=True, classes=set(classes))
visualizer.fit_transform(np.array(features), tclass_labels)
visualizer.poof(outpath="pca.png")
plt.close()
# os.system('open pca.png')
# spectral embedding
plt.figure()
viz = Manifold(manifold="spectral", classes=set(classes))
viz.fit_transform(np.array(features), tclass_labels)
viz.poof(outpath="spectral.png")
plt.close()
# lle embedding
plt.figure()
viz = Manifold(manifold="lle", classes=set(classes))
viz.fit_transform(np.array(features), tclass_labels)
viz.poof(outpath="lle.png")
plt.close()
# ltsa
# plt.figure()
# viz = Manifold(manifold="ltsa", classes=set(classes))
# viz.fit_transform(np.array(features), tclass_labels)
# viz.poof(outpath="ltsa.png")
# plt.close()
# hessian
# plt.figure()
# viz = Manifold(manifold="hessian", method='dense', classes=set(classes))
# viz.fit_transform(np.array(features), tclass_labels)
# viz.poof(outpath="hessian.png")
# plt.close()
# modified
plt.figure()
viz = Manifold(manifold="modified", classes=set(classes))
viz.fit_transform(np.array(features), tclass_labels)
viz.poof(outpath="modified.png")
plt.close()
# isomap
plt.figure()
viz = Manifold(manifold="isomap", classes=set(classes))
viz.fit_transform(np.array(features), tclass_labels)
viz.poof(outpath="isomap.png")
plt.close()
# mds
plt.figure()
viz = Manifold(manifold="mds", classes=set(classes))
viz.fit_transform(np.array(features), tclass_labels)
viz.poof(outpath="mds.png")
plt.close()
# spectral
plt.figure()
viz = Manifold(manifold="spectral", classes=set(classes))
viz.fit_transform(np.array(features), tclass_labels)
viz.poof(outpath="spectral.png")
plt.close()
# UMAP embedding
plt.figure()
umap = UMAPVisualizer(metric='cosine',
classes=set(classes),
title="UMAP embedding")
umap.fit_transform(np.array(features), class_labels)
umap.poof(outpath="umap.png")
plt.close()
# alternative UMAP
# import umap.plot
# plt.figure()
# mapper = umap.UMAP().fit(np.array(features))
# fig=umap.plot.points(mapper, labels=np.array(tclass_labels))
# fig = fig.get_figure()
# fig.tight_layout()
# fig.savefig('umap2.png')
# plt.close(fig)
#################################
# FEATURE RANKING!!
#################################
os.chdir(curdir)
os.mkdir('feature_ranking')
os.chdir('feature_ranking')
# You can get the feature importance of each feature of your dataset
# by using the feature importance property of the model.
plt.figure(figsize=(12, 12))
model = ExtraTreesClassifier()
model.fit(np.array(features), tclass_labels)
# print(model.feature_importances_)
feat_importances = pd.Series(model.feature_importances_,
index=feature_labels[0])
feat_importances.nlargest(20).plot(kind='barh')
plt.title('Feature importances (ExtraTrees)', size=16)
plt.title('Feature importances with %s features' % (str(len(features[0]))))
plt.tight_layout()
plt.savefig('feature_importance.png')
plt.close()
# os.system('open feature_importance.png')
# get selected labels for top 20 features
selectedlabels = list(dict(feat_importances.nlargest(20)))
new_features, new_labels = restructure_features(selectedlabels, t_features,
feature_labels[0])
new_features_, new_labels_ = restructure_features(selectedlabels, features,
feature_labels[0])
# Shapiro rank algorithm (1D)
plt.figure(figsize=(28, 12))
visualizer = Rank1D(algorithm='shapiro',
classes=set(classes),
features=new_labels)
visualizer.fit(np.array(new_features), tclass_labels)
visualizer.transform(np.array(new_features))
# plt.tight_layout()
visualizer.poof(outpath="shapiro.png")
plt.title('Shapiro plot (top 20 features)', size=16)
plt.close()
# os.system('open shapiro.png')
# visualizer.show()
# pearson ranking algorithm (2D)
plt.figure(figsize=(12, 12))
visualizer = Rank2D(algorithm='pearson',
classes=set(classes),
features=new_labels)
visualizer.fit(np.array(new_features), tclass_labels)
visualizer.transform(np.array(new_features))
plt.tight_layout()
visualizer.poof(outpath="pearson.png")
plt.title('Pearson ranking plot (top 20 features)', size=16)
plt.close()
# os.system('open pearson.png')
# visualizer.show()
# feature importances with top 20 features for Lasso
plt.figure(figsize=(12, 12))
viz = FeatureImportances(Lasso(), labels=new_labels_)
viz.fit(np.array(new_features_), tclass_labels)
plt.tight_layout()
viz.poof(outpath="lasso.png")
plt.close()
# correlation plots with feature removal if corr > 0.90
# https://towardsdatascience.com/feature-selection-correlation-and-p-value-da8921bfb3cf
# now remove correlated features
# --> p values
# --> https://towardsdatascience.com/the-next-level-of-data-visualization-in-python-dd6e99039d5e / https://github.com/WillKoehrsen/Data-Analysis/blob/master/plotly/Plotly%20Whirlwind%20Introduction.ipynb- plotly for correlation heatmap and scatterplot matrix
# --> https://seaborn.pydata.org/tutorial/distributions.html
data = new_features
corr = data.corr()
plt.figure(figsize=(12, 12))
fig = sns.heatmap(corr)
fig = fig.get_figure()
plt.title('Heatmap with correlated features (top 20 features)', size=16)
fig.tight_layout()
fig.savefig('heatmap.png')
plt.close(fig)
columns = np.full((corr.shape[0], ), True, dtype=bool)
for i in range(corr.shape[0]):
for j in range(i + 1, corr.shape[0]):
if corr.iloc[i, j] >= 0.9:
if columns[j]:
columns[j] = False
selected_columns = data.columns[columns]
data = data[selected_columns]
corr = data.corr()
plt.figure(figsize=(12, 12))
fig = sns.heatmap(corr)
fig = fig.get_figure()
plt.title('Heatmap without correlated features (top 20 features)', size=16)
fig.tight_layout()
fig.savefig('heatmap_clean.png')
plt.close(fig)
# radviz
# Instantiate the visualizer
plt.figure(figsize=(12, 12))
visualizer = RadViz(classes=classes, features=new_labels)
visualizer.fit(np.array(new_features), tclass_labels)
visualizer.transform(np.array(new_features))
visualizer.poof(outpath="radviz.png")
visualizer.show()
plt.close()
# feature correlation plot
plt.figure(figsize=(28, 12))
visualizer = feature_correlation(np.array(new_features),
tclass_labels,
labels=new_labels)
visualizer.poof(outpath="correlation.png")
visualizer.show()
plt.tight_layout()
plt.close()
os.mkdir('feature_plots')
os.chdir('feature_plots')
newdata = new_features_
newdata['classes'] = class_labels
for j in range(len(new_labels_)):
fig = sns.violinplot(x=newdata['classes'], y=newdata[new_labels_[j]])
fig = fig.get_figure()
fig.tight_layout()
fig.savefig('%s_%s.png' % (str(j), new_labels_[j]))
plt.close(fig)
os.mkdir('feature_plots_transformed')
os.chdir('feature_plots_transformed')
newdata = new_features
newdata['classes'] = class_labels
for j in range(len(new_labels)):
fig = sns.violinplot(x=newdata['classes'], y=newdata[new_labels[j]])
fig = fig.get_figure()
fig.tight_layout()
fig.savefig('%s_%s.png' % (str(j), new_labels[j]))
plt.close(fig)
##################################################
# PRECISION-RECALL CURVES
##################################################
os.chdir(curdir)
os.mkdir('model_selection')
os.chdir('model_selection')
plt.figure()
visualizer = precision_recall_curve(GaussianNB(), np.array(features),
tclass_labels)
visualizer.poof(outpath="precision-recall.png")
plt.close()
plt.figure()
visualizer = roc_auc(LogisticRegression(), np.array(features),
tclass_labels)
visualizer.poof(outpath="roc_curve_train.png")
plt.close()
plt.figure()
visualizer = discrimination_threshold(
LogisticRegression(multi_class="auto", solver="liblinear"),
np.array(features), tclass_labels)
visualizer.poof(outpath="thresholds.png")
plt.close()
plt.figure()
visualizer = residuals_plot(Ridge(),
np.array(features),
tclass_labels,
train_color="maroon",
test_color="gold")
visualizer.poof(outpath="residuals.png")
plt.close()
plt.figure()
visualizer = prediction_error(Lasso(), np.array(features), tclass_labels)
visualizer.poof(outpath='prediction_error.png')
plt.close()
# outlier detection
plt.figure()
visualizer = cooks_distance(np.array(features),
tclass_labels,
draw_threshold=True,
linefmt="C0-",
markerfmt=",")
visualizer.poof(outpath='outliers.png')
plt.close()
# cluster numbers
plt.figure()
visualizer = silhouette_visualizer(
KMeans(len(set(tclass_labels)), random_state=42), np.array(features))
visualizer.poof(outpath='siloutte.png')
plt.close()
# cluster distance
plt.figure()
visualizer = intercluster_distance(
KMeans(len(set(tclass_labels)), random_state=777), np.array(features))
visualizer.poof(outpath='cluster_distance.png')
plt.close()
# plot percentile of features plot with SVM to see which percentile for features is optimal
features = preprocessing.MinMaxScaler().fit_transform(features)
clf = Pipeline([('anova', SelectPercentile(chi2)),
('scaler', StandardScaler()),
('logr', LogisticRegression())])
score_means = list()
score_stds = list()
percentiles = (1, 3, 6, 10, 15, 20, 30, 40, 50, 60, 70, 80, 90, 100)
for percentile in percentiles:
clf.set_params(anova__percentile=percentile)
this_scores = cross_val_score(clf, np.array(features), class_labels)
score_means.append(this_scores.mean())
score_stds.append(this_scores.std())
plt.errorbar(percentiles, score_means, np.array(score_stds))
plt.title(
'Performance of the LogisticRegression-Anova varying the percent features selected'
)
plt.xticks(np.linspace(0, 100, 11, endpoint=True))
plt.xlabel('Percentile')
plt.ylabel('Accuracy Score')
plt.axis('tight')
plt.savefig('logr_percentile_plot.png')
plt.close()
# get PCA
pca = PCA(random_state=1)
pca.fit(X_train)
skplt.decomposition.plot_pca_component_variance(pca)
plt.savefig('pca_explained_variance.png')
plt.close()
# estimators
rf = RandomForestClassifier()
skplt.estimators.plot_learning_curve(rf, X_train, y_train)
plt.title('Learning Curve (Random Forest)')
plt.savefig('learning_curve.png')
plt.close()
# elbow plot
kmeans = KMeans(random_state=1)
skplt.cluster.plot_elbow_curve(kmeans,
X_train,
cluster_ranges=range(1, 30),
title='Elbow plot (KMeans clustering)')
plt.savefig('elbow.png')
plt.close()
# KS statistic (only if 2 classes)
lr = LogisticRegression()
lr = lr.fit(X_train, y_train)
y_probas = lr.predict_proba(X_test)
skplt.metrics.plot_ks_statistic(y_test, y_probas)
plt.savefig('ks.png')
plt.close()
# precision-recall
nb = GaussianNB()
nb.fit(X_train, y_train)
y_probas = nb.predict_proba(X_test)
skplt.metrics.plot_precision_recall(y_test, y_probas)
plt.tight_layout()
plt.savefig('precision-recall.png')
plt.close()
## plot calibration curve
rf = RandomForestClassifier()
lr = LogisticRegression()
nb = GaussianNB()
svm = LinearSVC()
dt = DecisionTreeClassifier(random_state=0)
ab = AdaBoostClassifier(n_estimators=100)
gb = GradientBoostingClassifier(n_estimators=100,
learning_rate=1.0,
max_depth=1,
random_state=0)
knn = KNeighborsClassifier(n_neighbors=7)
rf_probas = rf.fit(X_train, y_train).predict_proba(X_test)
lr_probas = lr.fit(X_train, y_train).predict_proba(X_test)
nb_probas = nb.fit(X_train, y_train).predict_proba(X_test)
# svm_scores = svm.fit(X_train, y_train).predict_proba(X_test)
dt_scores = dt.fit(X_train, y_train).predict_proba(X_test)
ab_scores = ab.fit(X_train, y_train).predict_proba(X_test)
gb_scores = gb.fit(X_train, y_train).predict_proba(X_test)
knn_scores = knn.fit(X_train, y_train).predict_proba(X_test)
probas_list = [
rf_probas,
lr_probas,
nb_probas, # svm_scores,
dt_scores,
ab_scores,
gb_scores,
knn_scores
]
clf_names = [
'Random Forest',
'Logistic Regression',
'Gaussian NB', # 'SVM',
'Decision Tree',
'Adaboost',
'Gradient Boost',
'KNN'
]
skplt.metrics.plot_calibration_curve(y_test, probas_list, clf_names)
plt.savefig('calibration.png')
plt.tight_layout()
plt.close()
# pick classifier type by ROC (without optimization)
probs = [
rf_probas[:, 1],
lr_probas[:, 1],
nb_probas[:, 1], # svm_scores[:, 1],
dt_scores[:, 1],
ab_scores[:, 1],
gb_scores[:, 1],
knn_scores[:, 1]
]
plot_roc_curve(y_test, probs, clf_names)
# more elaborate ROC example with CV = 5 fold
# https://scikit-learn.org/stable/auto_examples/model_selection/plot_roc_crossval.html#sphx-glr-auto-examples-model-selection-plot-roc-crossval-py
os.chdir(curdir)
return ''
def umap(
adata: AnnData,
min_dist: float = 0.5,
spread: float = 1.0,
n_components: int = 2,
maxiter: Optional[int] = None,
alpha: float = 1.0,
gamma: float = 1.0,
negative_sample_rate: int = 5,
init_pos: Union[_InitPos, np.ndarray, None] = 'spectral',
random_state: AnyRandom = 0,
a: Optional[float] = None,
b: Optional[float] = None,
copy: bool = False,
method: Literal['umap', 'rapids'] = 'umap',
neighbors_key: Optional[str] = None,
) -> Optional[AnnData]:
"""\
Embed the neighborhood graph using UMAP [McInnes18]_.
UMAP (Uniform Manifold Approximation and Projection) is a manifold learning
technique suitable for visualizing high-dimensional data. Besides tending to
be faster than tSNE, it optimizes the embedding such that it best reflects
the topology of the data, which we represent throughout Scanpy using a
neighborhood graph. tSNE, by contrast, optimizes the distribution of
nearest-neighbor distances in the embedding such that these best match the
distribution of distances in the high-dimensional space. We use the
implementation of `umap-learn <https://github.com/lmcinnes/umap>`__
[McInnes18]_. For a few comparisons of UMAP with tSNE, see this `preprint
<https://doi.org/10.1101/298430>`__.
Parameters
----------
adata
Annotated data matrix.
min_dist
The effective minimum distance between embedded points. Smaller values
will result in a more clustered/clumped embedding where nearby points on
the manifold are drawn closer together, while larger values will result
on a more even dispersal of points. The value should be set relative to
the ``spread`` value, which determines the scale at which embedded
points will be spread out. The default of in the `umap-learn` package is
0.1.
spread
The effective scale of embedded points. In combination with `min_dist`
this determines how clustered/clumped the embedded points are.
n_components
The number of dimensions of the embedding.
maxiter
The number of iterations (epochs) of the optimization. Called `n_epochs`
in the original UMAP.
alpha
The initial learning rate for the embedding optimization.
gamma
Weighting applied to negative samples in low dimensional embedding
optimization. Values higher than one will result in greater weight
being given to negative samples.
negative_sample_rate
The number of negative edge/1-simplex samples to use per positive
edge/1-simplex sample in optimizing the low dimensional embedding.
init_pos
How to initialize the low dimensional embedding. Called `init` in the
original UMAP. Options are:
* Any key for `adata.obsm`.
* 'paga': positions from :func:`~scanpy.pl.paga`.
* 'spectral': use a spectral embedding of the graph.
* 'random': assign initial embedding positions at random.
* A numpy array of initial embedding positions.
random_state
If `int`, `random_state` is the seed used by the random number generator;
If `RandomState` or `Generator`, `random_state` is the random number generator;
If `None`, the random number generator is the `RandomState` instance used
by `np.random`.
a
More specific parameters controlling the embedding. If `None` these
values are set automatically as determined by `min_dist` and
`spread`.
b
More specific parameters controlling the embedding. If `None` these
values are set automatically as determined by `min_dist` and
`spread`.
copy
Return a copy instead of writing to adata.
method
Use the original 'umap' implementation, or 'rapids' (experimental, GPU only)
neighbors_key
If not specified, umap looks .uns['neighbors'] for neighbors settings
and .obsp['connectivities'] for connectivities
(default storage places for pp.neighbors).
If specified, umap looks .uns[neighbors_key] for neighbors settings and
.obsp[.uns[neighbors_key]['connectivities_key']] for connectivities.
Returns
-------
Depending on `copy`, returns or updates `adata` with the following fields.
**X_umap** : `adata.obsm` field
UMAP coordinates of data.
"""
adata = adata.copy() if copy else adata
if neighbors_key is None:
neighbors_key = 'neighbors'
if neighbors_key not in adata.uns:
raise ValueError(
f'Did not find .uns["{neighbors_key}"]. Run `sc.pp.neighbors` first.'
)
start = logg.info('computing UMAP')
neighbors = NeighborsView(adata, neighbors_key)
if 'params' not in neighbors or neighbors['params']['method'] != 'umap':
logg.warning(
f'.obsp["{neighbors["connectivities_key"]}"] have not been computed using umap'
)
# Compat for umap 0.4 -> 0.5
with warnings.catch_warnings():
# umap 0.5.0
warnings.filterwarnings("ignore", message=r"Tensorflow not installed")
import umap
if version.parse(umap.__version__) >= version.parse("0.5.0"):
def simplicial_set_embedding(*args, **kwargs):
from umap.umap_ import simplicial_set_embedding
X_umap, _ = simplicial_set_embedding(
*args,
densmap=False,
densmap_kwds={},
output_dens=False,
**kwargs,
)
return X_umap
else:
from umap.umap_ import simplicial_set_embedding
from umap.umap_ import find_ab_params
if a is None or b is None:
a, b = find_ab_params(spread, min_dist)
else:
a = a
b = b
adata.uns['umap'] = {'params': {'a': a, 'b': b}}
if isinstance(init_pos, str) and init_pos in adata.obsm.keys():
init_coords = adata.obsm[init_pos]
elif isinstance(init_pos, str) and init_pos == 'paga':
init_coords = get_init_pos_from_paga(
adata, random_state=random_state, neighbors_key=neighbors_key
)
else:
init_coords = init_pos # Let umap handle it
if hasattr(init_coords, "dtype"):
init_coords = check_array(init_coords, dtype=np.float32, accept_sparse=False)
if random_state != 0:
adata.uns['umap']['params']['random_state'] = random_state
random_state = check_random_state(random_state)
neigh_params = neighbors['params']
X = _choose_representation(
adata,
neigh_params.get('use_rep', None),
neigh_params.get('n_pcs', None),
silent=True,
)
if method == 'umap':
# the data matrix X is really only used for determining the number of connected components
# for the init condition in the UMAP embedding
n_epochs = 0 if maxiter is None else maxiter
X_umap = simplicial_set_embedding(
X,
neighbors['connectivities'].tocoo(),
n_components,
alpha,
a,
b,
gamma,
negative_sample_rate,
n_epochs,
init_coords,
random_state,
neigh_params.get('metric', 'euclidean'),
neigh_params.get('metric_kwds', {}),
verbose=settings.verbosity > 3,
)
elif method == 'rapids':
metric = neigh_params.get('metric', 'euclidean')
if metric != 'euclidean':
raise ValueError(
f'`sc.pp.neighbors` was called with `metric` {metric!r}, '
"but umap `method` 'rapids' only supports the 'euclidean' metric."
)
from cuml import UMAP
n_neighbors = neighbors['params']['n_neighbors']
n_epochs = (
500 if maxiter is None else maxiter
) # 0 is not a valid value for rapids, unlike original umap
X_contiguous = np.ascontiguousarray(X, dtype=np.float32)
umap = UMAP(
n_neighbors=n_neighbors,
n_components=n_components,
n_epochs=n_epochs,
learning_rate=alpha,
init=init_pos,
min_dist=min_dist,
spread=spread,
negative_sample_rate=negative_sample_rate,
a=a,
b=b,
verbose=settings.verbosity > 3,
random_state=random_state,
)
X_umap = umap.fit_transform(X_contiguous)
adata.obsm['X_umap'] = X_umap # annotate samples with UMAP coordinates
logg.info(
' finished',
time=start,
deep=('added\n' " 'X_umap', UMAP coordinates (adata.obsm)"),
)
return adata if copy else None
#esto ya no podr'a ser visualizado en 2D porque tiene 3 atributos #entonces vamos a generar una proyecci'on 2D de estos datos X = dataset.iloc[:, [2, 3, 4]].values X = StandardScaler().fit_transform(X) #vamos a usar una proyeccion llamada UMAP es #reciente y mostr'o resultados buen'isimos umap = umap.UMAP(n_neighbors=3, min_dist=0.6, metric='cosine') #esta proyeccion solo va a ser usada para visualizaciones #nuestro clustering ser'a hecho en el espacio ndimensional #que en este caso n=3 X_projected = umap.fit_transform(X) plt.scatter(X_projected[:, 0], X_projected[:, 1], s=100, color="blue") plt.grid() plt.show() kmeans = KMeans(n_clusters=7, init="random", max_iter=300) # dentro de pred_y va a tener la lista de clusters resultantes # NOO es variable dependiente, es un atributo nuevo # son las etiquetas de cluster/grupo predichas labels = kmeans.fit_predict(X) plt.scatter(X_projected[:, 0], X_projected[:, 1], s=100, c=labels) plt.grid() # una opcion para mostrar nuestra malla grafica #esta es la posicion de cada centroide en el grafico # que no necesiramente esta asociado a un dato