def ex3(wv):
phrases = [
"szkoda",
"strata",
"uszczerbek",
"szkoda majątkowa",
# "uszczerbek na zdrowiu",
"krzywda",
"niesprawiedliwość",
"nieszczęście"
]
tsne = MulticoreTSNE(n_components=2, n_jobs=os.cpu_count())
tsne.fit(wv.vectors)
vectors_embedded = tsne.fit_transform(wv[(sanitize(phrase)
for phrase in phrases)])
fig, ax = plt.subplots()
ax.scatter(vectors_embedded[:, 0], vectors_embedded[:, 1])
for i, phrase in enumerate(phrases):
ax.annotate(phrase, (vectors_embedded[:, 0][i], vectors_embedded[:, 1][i]))
plt.show()
def toTsne(train,test,n_component=2,file_name='tf_idf',path='data/'):
tsne=TSNE(n_components=n_component,random_state=1123,njobs=-1)
lentrain=train.shape[0]
X=np.vstack([train,test])
tsne.fit(X)
res=tsne.embedding_
#print res
pd.to_pickle(res[:lentrain],path+'train_svd_20_tsne_'+str(n_component)+'_'+file_name+'.pkl')
pd.to_pickle(res[lentrain:],path+'test_svd_20_tsne_'+str(n_component)+'_'+file_name+'.pkl')
return 'Success'
def tsneError_MultiCpore_Tsne(params, data, perplexity):
# need to not recreate P
mtSNE = MulticoreTSNE(perplexity=perplexity,
init=params,
n_iter=0,
early_exaggeration=1)
mtSNE = mtSNE.fit(data)
error = mtSNE.kl_divergence_
print(error)
# Yeah it's a hack.
return error
def fit_tsne(X,
perplexity=PERPLEXITY,
n_iter=N_ITER,
learn_rate=200.,
early_exag=12.):
try:
from MulticoreTSNE import MulticoreTSNE
tsne = MulticoreTSNE(n_iter=500,
perplexity=perplexity,
learning_rate=learn_rate,
early_exaggeration=early_exag,
random_state=69,
n_jobs=40)
except ImportError:
tsne = TSNEApprox(
n_iter=500,
perplexity=perplexity,
learning_rate=learn_rate,
early_exaggeration=early_exag,
random_state=69,
)
tsne.fit(X)
embedding = tsne.embedding_
return embedding
def visualize(assembled,
labels,
namespace,
data_names,
gene_names=None,
gene_expr=None,
genes=None,
n_iter=N_ITER,
perplexity=PERPLEXITY,
verbose=VERBOSE,
learn_rate=200.,
early_exag=12.,
embedding=None,
shuffle_ds=False,
size=1,
multicore_tsne=True,
image_suffix='.svg',
viz_cluster=False,
colors=None):
# Fit t-SNE.
if embedding is None:
try:
from MulticoreTSNE import MulticoreTSNE
tsne = MulticoreTSNE(n_iter=n_iter,
perplexity=perplexity,
verbose=verbose,
random_state=69,
learning_rate=learn_rate,
early_exaggeration=early_exag,
n_jobs=40)
except ImportError:
multicore_tsne = False
if not multicore_tsne:
tsne = TSNEApprox(n_iter=n_iter,
perplexity=perplexity,
verbose=verbose,
random_state=69,
learning_rate=learn_rate,
early_exaggeration=early_exag)
tsne.fit(np.concatenate(assembled))
embedding = tsne.embedding_
if shuffle_ds:
rand_idx = range(embedding.shape[0])
random.shuffle(list(rand_idx))
embedding = embedding[rand_idx, :]
labels = labels[rand_idx]
# Plot clusters together.
plot_clusters(embedding, labels, s=size, colors=colors)
plt.title(('Panorama ({} iter, perplexity: {}, sigma: {}, ' +
'knn: {}, hvg: {}, dimred: {}, approx: {})').format(
n_iter, perplexity, SIGMA, KNN, HVG, DIMRED, APPROX))
plt.savefig(namespace + image_suffix, dpi=500)
# Plot clusters individually.
if viz_cluster and not shuffle_ds:
for i in range(len(data_names)):
visualize_cluster(embedding,
i,
labels,
cluster_name=data_names[i],
size=size,
viz_prefix=namespace,
image_suffix=image_suffix)
# Plot gene expression levels.
if (not gene_names is None) and \
(not gene_expr is None) and \
(not genes is None):
if shuffle_ds:
gene_expr = gene_expr[rand_idx, :]
for gene_name in gene_names:
visualize_expr(gene_expr,
embedding,
genes,
gene_name,
size=size,
viz_prefix=namespace,
image_suffix=image_suffix)
return embedding