Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit cff8a4a

Browse files
ArturoAmorQogrisel
authored and
jeremiedbb
committed
DOC Change print format in TSNE example (#25569)
Co-authored-by: Olivier Grisel <[email protected]>
1 parent e90002a commit cff8a4a

File tree

1 file changed

+30
-35
lines changed

1 file changed

+30
-35
lines changed

‎examples/neighbors/approximate_nearest_neighbors.py

Lines changed: 30 additions & 35 deletions
Original file line numberDiff line numberDiff line change
@@ -156,54 +156,50 @@ def load_mnist(n_samples):
156156

157157
for dataset_name, (X, y) in datasets:
158158

159-
msg = "Benchmarking on %s:" % dataset_name
160-
print("\n%s\n%s" % (msg, "-" * len(msg)))
159+
msg = f"Benchmarking on {dataset_name}:"
160+
print(f"\n{msg}\n" + str("-" * len(msg)))
161161

162162
for transformer_name, transformer in transformers:
163163
longest = np.max([len(name) for name, model in transformers])
164-
whitespaces = " " * (longest - len(transformer_name))
165164
start = time.time()
166165
transformer.fit(X)
167166
fit_duration = time.time() - start
168-
print("%s: %s%.3f sec (fit)" % (transformer_name, whitespaces, fit_duration))
167+
print(f"{transformer_name:<{longest}} {fit_duration:.3f} sec (fit)")
169168
start = time.time()
170169
Xt = transformer.transform(X)
171170
transform_duration = time.time() - start
172-
print(
173-
"%s: %s%.3f sec (transform)"
174-
% (transformer_name, whitespaces, transform_duration)
175-
)
171+
print(f"{transformer_name:<{longest}} {transform_duration:.3f} sec (transform)")
176172
if transformer_name == "PyNNDescentTransformer":
177173
start = time.time()
178174
Xt = transformer.transform(X)
179175
transform_duration = time.time() - start
180176
print(
181-
"%s: %s%.3f sec (transform)"
182-
% (transformer_name, whitespaces, transform_duration)
177+
f"{transformer_name:<{longest}} {transform_duration:.3f} sec"
178+
" (transform)"
183179
)
184180

185181
# %%
186182
# Sample output::
187183
#
188184
# Benchmarking on MNIST_10000:
189185
# ----------------------------
190-
# KNeighborsTransformer: 0.007 sec (fit)
191-
# KNeighborsTransformer: 1.139 sec (transform)
192-
# NMSlibTransformer: 0.208 sec (fit)
193-
# NMSlibTransformer: 0.315 sec (transform)
194-
# PyNNDescentTransformer: 4.823 sec (fit)
195-
# PyNNDescentTransformer: 4.884 sec (transform)
196-
# PyNNDescentTransformer: 0.744 sec (transform)
186+
# KNeighborsTransformer 0.007 sec (fit)
187+
# KNeighborsTransformer 1.139 sec (transform)
188+
# NMSlibTransformer 0.208 sec (fit)
189+
# NMSlibTransformer 0.315 sec (transform)
190+
# PyNNDescentTransformer 4.823 sec (fit)
191+
# PyNNDescentTransformer 4.884 sec (transform)
192+
# PyNNDescentTransformer 0.744 sec (transform)
197193
#
198194
# Benchmarking on MNIST_20000:
199195
# ----------------------------
200-
# KNeighborsTransformer: 0.011 sec (fit)
201-
# KNeighborsTransformer: 5.769 sec (transform)
202-
# NMSlibTransformer: 0.733 sec (fit)
203-
# NMSlibTransformer: 1.077 sec (transform)
204-
# PyNNDescentTransformer: 14.448 sec (fit)
205-
# PyNNDescentTransformer: 7.103 sec (transform)
206-
# PyNNDescentTransformer: 1.759 sec (transform)
196+
# KNeighborsTransformer 0.011 sec (fit)
197+
# KNeighborsTransformer 5.769 sec (transform)
198+
# NMSlibTransformer 0.733 sec (fit)
199+
# NMSlibTransformer 1.077 sec (transform)
200+
# PyNNDescentTransformer 14.448 sec (fit)
201+
# PyNNDescentTransformer 7.103 sec (transform)
202+
# PyNNDescentTransformer 1.759 sec (transform)
207203
#
208204
# Notice that the `PyNNDescentTransformer` takes more time during the first
209205
# `fit` and the first `transform` due to the overhead of the numba just in time
@@ -248,18 +244,17 @@ def load_mnist(n_samples):
248244

249245
for dataset_name, (X, y) in datasets:
250246

251-
msg = "Benchmarking on %s:" % dataset_name
252-
print("\n%s\n%s" % (msg, "-" * len(msg)))
247+
msg = f"Benchmarking on {dataset_name}:"
248+
print(f"\n{msg}\n" + str("-" * len(msg)))
253249

254250
for transformer_name, transformer in transformers:
255251
longest = np.max([len(name) for name, model in transformers])
256-
whitespaces = " " * (longest - len(transformer_name))
257252
start = time.time()
258253
Xt = transformer.fit_transform(X)
259254
transform_duration = time.time() - start
260255
print(
261-
"%s: %s%.3f sec (fit_transform)"
262-
% (transformer_name, whitespaces, transform_duration)
256+
f"{transformer_name:<{longest}} {transform_duration:.3f} sec"
257+
" (fit_transform)"
263258
)
264259

265260
# plot TSNE embedding which should be very similar across methods
@@ -284,15 +279,15 @@ def load_mnist(n_samples):
284279
#
285280
# Benchmarking on MNIST_10000:
286281
# ----------------------------
287-
# TSNE with internal NearestNeighbors: 24.828 sec (fit_transform)
288-
# TSNE with KNeighborsTransformer: 20.111 sec (fit_transform)
289-
# TSNE with NMSlibTransformer: 21.757 sec (fit_transform)
282+
# TSNE with internal NearestNeighbors 24.828 sec (fit_transform)
283+
# TSNE with KNeighborsTransformer 20.111 sec (fit_transform)
284+
# TSNE with NMSlibTransformer 21.757 sec (fit_transform)
290285
#
291286
# Benchmarking on MNIST_20000:
292287
# ----------------------------
293-
# TSNE with internal NearestNeighbors: 51.955 sec (fit_transform)
294-
# TSNE with KNeighborsTransformer: 50.994 sec (fit_transform)
295-
# TSNE with NMSlibTransformer: 43.536 sec (fit_transform)
288+
# TSNE with internal NearestNeighbors 51.955 sec (fit_transform)
289+
# TSNE with KNeighborsTransformer 50.994 sec (fit_transform)
290+
# TSNE with NMSlibTransformer 43.536 sec (fit_transform)
296291
#
297292
# We can observe that the default :class:`~sklearn.manifold.TSNE` estimator with
298293
# its internal :class:`~sklearn.neighbors.NearestNeighbors` implementation is

0 commit comments

Comments
 (0)