ENH lots of benchmarks fixes

* add a title to all benchmark figures/windows
* "bench" is non-standard slang for benchmarking
* remove calls to pyplot.clf because the previous call
  to pyplot.figure sets the current figure
* standardize on "Time (s)" as time axis label
* move some benchmark plot legends to the upper left corner of the plot:
  most benchmarks expand from (0,0) up and to the right
* simplified plot titles and moved constant info from legend into title
* fixed typos
* made some labels more viewer-friendly

Fixes scikit-learn#2023.

benchmarks/bench_glm.py

@@ -46,12 +46,13 @@
         lasso.fit(X, Y)
         time_lasso[i] = total_seconds(datetime.now() - start)
 
-    pl.xlabel('Dimesions')
-    pl.ylabel('Time (in seconds)')
+    pl.figure('scikit-learn GLM benchmark results')
+    pl.xlabel('Dimensions')
+    pl.ylabel('Time (s)')
     pl.plot(dimensions, time_ridge, color='r')
     pl.plot(dimensions, time_ols, color='g')
     pl.plot(dimensions, time_lasso, color='b')
 
-    pl.legend(['Ridge', 'OLS', 'LassoLars'])
+    pl.legend(['Ridge', 'OLS', 'LassoLars'], loc='upper left')
     pl.axis('tight')
     pl.show()

benchmarks/bench_glmnet.py

@@ -71,9 +71,9 @@ def bench(factory, X, Y, X_test, Y_test, ref_coef):
         X = X[:(i * step)]
         Y = Y[:(i * step)]
 
-        print("benching scikit-learn: ")
+        print("benchmarking scikit-learn: ")
         scikit_results.append(bench(ScikitLasso, X, Y, X_test, Y_test, coef_))
-        print("benching glmnet: ")
+        print("benchmarking glmnet: ")
         glmnet_results.append(bench(GlmnetLasso, X, Y, X_test, Y_test, coef_))
 
     pl.clf()
@@ -83,10 +83,10 @@ def bench(factory, X, Y, X_test, Y_test, ref_coef):
     pl.plot(xx, glmnet_results, 'r-', label='glmnet')
     pl.legend()
     pl.xlabel('number of samples to classify')
-    pl.ylabel('time (in seconds)')
+    pl.ylabel('Time (s)')
     pl.show()
 
-    # now do a bench where the number of points is fixed
+    # now do a benchmark where the number of points is fixed
     # and the variable is the number of features
 
     scikit_results = []
@@ -111,18 +111,18 @@ def bench(factory, X, Y, X_test, Y_test, ref_coef):
         X = X[:n_samples]
         Y = Y[:n_samples]
 
-        print("benching scikit-learn: ")
+        print("benchmarking scikit-learn: ")
         scikit_results.append(bench(ScikitLasso, X, Y, X_test, Y_test, coef_))
-        print("benching glmnet: ")
+        print("benchmarking glmnet: ")
         glmnet_results.append(bench(GlmnetLasso, X, Y, X_test, Y_test, coef_))
 
     xx = np.arange(100, 100 + n * step, step)
-    pl.figure()
+    pl.figure('scikit-learn vs. glmnet benchmark results')
     pl.title('Regression in high dimensional spaces (%d samples)' % n_samples)
     pl.plot(xx, scikit_results, 'b-', label='scikit-learn')
     pl.plot(xx, glmnet_results, 'r-', label='glmnet')
     pl.legend()
     pl.xlabel('number of features')
-    pl.ylabel('time (in seconds)')
+    pl.ylabel('Time (s)')
     pl.axis('tight')
     pl.show()

benchmarks/bench_lasso.py

@@ -39,15 +39,15 @@ def compute_bench(alpha, n_samples, n_features, precompute):
             X /= np.sqrt(np.sum(X ** 2, axis=0))  # Normalize data
 
             gc.collect()
-            print("- benching Lasso")
+            print("- benchmarking Lasso")
             clf = Lasso(alpha=alpha, fit_intercept=False,
                         precompute=precompute)
             tstart = time()
             clf.fit(X, Y)
             lasso_results.append(time() - tstart)
 
             gc.collect()
-            print("- benching LassoLars")
+            print("- benchmarking LassoLars")
             clf = LassoLars(alpha=alpha, fit_intercept=False,
                             normalize=False, precompute=precompute)
             tstart = time()
@@ -68,16 +68,16 @@ def compute_bench(alpha, n_samples, n_features, precompute):
     lasso_results, lars_lasso_results = compute_bench(alpha, list_n_samples,
                                             [n_features], precompute=True)
 
-    pl.clf()
+    pl.figure('scikit-learn LASSO benchmark results')
     pl.subplot(211)
     pl.plot(list_n_samples, lasso_results, 'b-',
-                            label='Lasso (with precomputed Gram matrix)')
+                            label='Lasso')
     pl.plot(list_n_samples, lars_lasso_results, 'r-',
-                            label='LassoLars (with precomputed Gram matrix)')
-    pl.title('Lasso benchmark (%d features - alpha=%s)' % (n_features, alpha))
+                            label='LassoLars')
+    pl.title('precomputed Gram matrix, %d features, alpha=%s' % (n_features, alpha))
     pl.legend(loc='upper left')
     pl.xlabel('number of samples')
-    pl.ylabel('time (in seconds)')
+    pl.ylabel('Time (s)')
     pl.axis('tight')
 
     n_samples = 2000
@@ -87,9 +87,9 @@ def compute_bench(alpha, n_samples, n_features, precompute):
     pl.subplot(212)
     pl.plot(list_n_features, lasso_results, 'b-', label='Lasso')
     pl.plot(list_n_features, lars_lasso_results, 'r-', label='LassoLars')
-    pl.title('Lasso benchmark (%d samples - alpha=%s)' % (n_samples, alpha))
+    pl.title('%d samples, alpha=%s' % (n_samples, alpha))
     pl.legend(loc='upper left')
     pl.xlabel('number of features')
-    pl.ylabel('time (in seconds)')
+    pl.ylabel('Time (s)')
     pl.axis('tight')
     pl.show()

benchmarks/bench_plot_fastkmeans.py

@@ -50,8 +50,8 @@ def compute_bench(samples_range, features_range):
             print()
             print()
 
-            results['minibatchkmeans_speed'].append(delta)
-            results['minibatchkmeans_quality'].append(mbkmeans.inertia_)
+            results['MiniBatchKMeans Speed'].append(delta)
+            results['MiniBatchKMeans Quality'].append(mbkmeans.inertia_)
 
     return results
 
@@ -85,8 +85,8 @@ def compute_bench_2(chunks):
         print("Inertia: %0.3fs" % mbkmeans.inertia_)
         print()
 
-        results['minibatchkmeans_speed'].append(delta)
-        results['minibatchkmeans_quality'].append(mbkmeans.inertia_)
+        results['MiniBatchKMeans Speed'].append(delta)
+        results['MiniBatchKMeans Quality'].append(mbkmeans.inertia_)
 
     return results
 
@@ -108,7 +108,7 @@ def compute_bench_2(chunks):
                         t for (label, t) in results.iteritems()
                             if "speed" not in label]])
 
-    fig = plt.figure()
+    fig = plt.figure('scikit-learn K-Means benchmark results')
     for c, (label, timings) in zip('brcy',
                                     sorted(results.iteritems())):
         if 'speed' in label:
@@ -132,7 +132,7 @@ def compute_bench_2(chunks):
         ax = fig.add_subplot(2, 2, i + 2)
         y = np.asarray(timings)
         ax.plot(chunks, y, color=c, alpha=0.8)
-        ax.set_xlabel('chunks')
+        ax.set_xlabel('Chunks')
         ax.set_ylabel(label)
 
     plt.show()

benchmarks/bench_plot_lasso_path.py

@@ -42,7 +42,7 @@ def compute_bench(samples_range, features_range):
             X, y = make_regression(**dataset_kwargs)
 
             gc.collect()
-            print("benching lars_path (with Gram):", end='')
+            print("benchmarking lars_path (with Gram):", end='')
             sys.stdout.flush()
             tstart = time()
             G = np.dot(X.T, X)  # precomputed Gram matrix
@@ -53,7 +53,7 @@ def compute_bench(samples_range, features_range):
             results['lars_path (with Gram)'].append(delta)
 
             gc.collect()
-            print("benching lars_path (without Gram):", end='')
+            print("benchmarking lars_path (without Gram):", end='')
             sys.stdout.flush()
             tstart = time()
             lars_path(X, y, method='lasso')
@@ -62,7 +62,7 @@ def compute_bench(samples_range, features_range):
             results['lars_path (without Gram)'].append(delta)
 
             gc.collect()
-            print("benching lasso_path (with Gram):", end='')
+            print("benchmarking lasso_path (with Gram):", end='')
             sys.stdout.flush()
             tstart = time()
             lasso_path(X, y, precompute=True)
@@ -71,7 +71,7 @@ def compute_bench(samples_range, features_range):
             results['lasso_path (with Gram)'].append(delta)
 
             gc.collect()
-            print("benching lasso_path (without Gram):", end='')
+            print("benchmarking lasso_path (without Gram):", end='')
             sys.stdout.flush()
             tstart = time()
             lasso_path(X, y, precompute=False)
@@ -92,7 +92,7 @@ def compute_bench(samples_range, features_range):
 
     max_time = max(max(t) for t in results.values())
 
-    fig = plt.figure()
+    fig = plt.figure('scikit-learn Lasso path benchmark results')
     i = 1
     for c, (label, timings) in zip('bcry', sorted(results.items())):
         ax = fig.add_subplot(2, 2, i, projection='3d')
@@ -109,7 +109,7 @@ def compute_bench(samples_range, features_range):
 
         ax.set_xlabel('n_samples')
         ax.set_ylabel('n_features')
-        ax.set_zlabel('time (s)')
+        ax.set_zlabel('Time (s)')
         ax.set_zlim3d(0.0, max_time * 1.1)
         ax.set_title(label)
         #ax.legend()

benchmarks/bench_plot_neighbors.py

@@ -106,7 +106,8 @@ def barplot_neighbors(Nrange=2 ** np.arange(1, 11),
             k_results_build[algorithm][i] = (t1 - t0)
             k_results_query[algorithm][i] = (t2 - t1)
 
-    pl.figure(figsize=(8, 11))
+    pl.figure('scikit-learn nearest neighbors benchmark results',
+              figsize=(8, 11))
 
     for (sbplt, vals, quantity,
          build_time, query_time) in [(311, Nrange, 'N',
@@ -131,10 +132,10 @@ def barplot_neighbors(Nrange=2 ** np.arange(1, 11),
             xvals = 0.1 + i * (1 + len(vals)) + np.arange(len(vals))
             width = 0.8
 
-            pl.bar(xvals, build_time[alg] - bottom,
-                   width, bottom, color='r')
-            pl.bar(xvals, query_time[alg],
-                   width, build_time[alg], color='b')
+            c_bar = pl.bar(xvals, build_time[alg] - bottom,
+                           width, bottom, color='r')
+            q_bar = pl.bar(xvals, query_time[alg],
+                           width, build_time[alg], color='b')
 
             tick_vals += list(xvals + 0.5 * width)
             tick_labels += ['%i' % val for val in vals]
@@ -145,7 +146,7 @@ def barplot_neighbors(Nrange=2 ** np.arange(1, 11),
                     va='top',
                     bbox=dict(facecolor='w', edgecolor='w', alpha=0.5))
 
-            pl.ylabel('time (seconds)')
+            pl.ylabel('Time (s)')
 
         ax.xaxis.set_major_locator(ticker.FixedLocator(tick_vals))
         ax.xaxis.set_major_formatter(ticker.FixedFormatter(tick_labels))
@@ -174,10 +175,10 @@ def barplot_neighbors(Nrange=2 ** np.arange(1, 11),
                 transform=ax.transAxes, rotation=-90,
                 ha='right', va='center')
 
-        pl.gcf().suptitle("%s data\nred = construction;  blue = N-point query"
-                          % (dataset[0].upper() + dataset[1:]),
-                          fontsize=16)
+        pl.gcf().suptitle("%s data set" % dataset.capitalize(), fontsize=16)
 
+    pl.figlegend((c_bar, q_bar), ('construction', 'N-point query'),
+                 'upper right')
 
 if __name__ == '__main__':
     barplot_neighbors(dataset='digits')

benchmarks/bench_plot_nmf.py

@@ -85,7 +85,7 @@ def compute_bench(samples_range, features_range, rank=50, tolerance=1e-7):
                        effective_rank=rank,  tail_strength=0.2))
 
             gc.collect()
-            print("benching nndsvd-nmf: ")
+            print("benchmarking nndsvd-nmf: ")
             tstart = time()
             m = NMF(n_components=30, tol=tolerance, init='nndsvd').fit(X)
             tend = time() - tstart
@@ -94,7 +94,7 @@ def compute_bench(samples_range, features_range, rank=50, tolerance=1e-7):
             print(m.reconstruction_err_, tend)
 
             gc.collect()
-            print("benching nndsvda-nmf: ")
+            print("benchmarking nndsvda-nmf: ")
             tstart = time()
             m = NMF(n_components=30, init='nndsvda',
                     tol=tolerance).fit(X)
@@ -104,7 +104,7 @@ def compute_bench(samples_range, features_range, rank=50, tolerance=1e-7):
             print(m.reconstruction_err_, tend)
 
             gc.collect()
-            print("benching nndsvdar-nmf: ")
+            print("benchmarking nndsvdar-nmf: ")
             tstart = time()
             m = NMF(n_components=30, init='nndsvdar',
                     tol=tolerance).fit(X)
@@ -114,7 +114,7 @@ def compute_bench(samples_range, features_range, rank=50, tolerance=1e-7):
             print(m.reconstruction_err_, tend)
 
             gc.collect()
-            print("benching random-nmf")
+            print("benchmarking random-nmf")
             tstart = time()
             m = NMF(n_components=30, init=None, max_iter=1000,
                     tol=tolerance).fit(X)
@@ -124,7 +124,7 @@ def compute_bench(samples_range, features_range, rank=50, tolerance=1e-7):
             print(m.reconstruction_err_, tend)
 
             gc.collect()
-            print("benching alt-random-nmf")
+            print("benchmarking alt-random-nmf")
             tstart = time()
             W, H = alt_nnmf(X, r=30, R=None, tol=tolerance)
             tend = time() - tstart
@@ -145,7 +145,7 @@ def compute_bench(samples_range, features_range, rank=50, tolerance=1e-7):
     timeset, err = compute_bench(samples_range, features_range)
 
     for i, results in enumerate((timeset, err)):
-        fig = plt.figure()
+        fig = plt.figure('scikit-learn Non-Negative Matrix Factorization benchmkar results')
         ax = fig.gca(projection='3d')
         for c, (label, timings) in zip('rbgcm', sorted(results.iteritems())):
             X, Y = np.meshgrid(samples_range, features_range)
@@ -160,7 +160,7 @@ def compute_bench(samples_range, features_range, rank=50, tolerance=1e-7):
 
         ax.set_xlabel('n_samples')
         ax.set_ylabel('n_features')
-        zlabel = 'time (s)' if i == 0 else 'reconstruction error'
+        zlabel = 'Time (s)' if i == 0 else 'reconstruction error'
         ax.set_zlabel(zlabel)
         ax.legend()
         plt.show()

benchmarks/bench_plot_omp_lars.py

@@ -55,7 +55,7 @@ def compute_bench(samples_range, features_range):
             X = np.asfortranarray(X)
 
             gc.collect()
-            print("benching lars_path (with Gram):", end='')
+            print("benchmarking lars_path (with Gram):", end='')
             sys.stdout.flush()
             tstart = time()
             G = np.dot(X.T, X)  # precomputed Gram matrix
@@ -66,7 +66,7 @@ def compute_bench(samples_range, features_range):
             lars_gram[i_f, i_s] = delta
 
             gc.collect()
-            print("benching lars_path (without Gram):", end='')
+            print("benchmarking lars_path (without Gram):", end='')
             sys.stdout.flush()
             tstart = time()
             lars_path(X, y, Gram=None, max_iter=n_informative)
@@ -75,7 +75,7 @@ def compute_bench(samples_range, features_range):
             lars[i_f, i_s] = delta
 
             gc.collect()
-            print("benching orthogonal_mp (with Gram):", end='')
+            print("benchmarking orthogonal_mp (with Gram):", end='')
             sys.stdout.flush()
             tstart = time()
             orthogonal_mp(X, y, precompute_gram=True,
@@ -85,7 +85,7 @@ def compute_bench(samples_range, features_range):
             omp_gram[i_f, i_s] = delta
 
             gc.collect()
-            print("benching orthogonal_mp (without Gram):", end='')
+            print("benchmarking orthogonal_mp (without Gram):", end='')
             sys.stdout.flush()
             tstart = time()
             orthogonal_mp(X, y, precompute_gram=False,
@@ -106,7 +106,7 @@ def compute_bench(samples_range, features_range):
     max_time = max(np.max(t) for t in results.values())
 
     import pylab as pl
-    fig = pl.figure()
+    fig = pl.figure('scikit-learn OMP vs. LARS benchmark results')
     for i, (label, timings) in enumerate(sorted(results.iteritems())):
         ax = fig.add_subplot(1, 2, i)
         vmax = max(1 - timings.min(), -1 + timings.max())

benchmarks/bench_plot_parallel_pairwise.py

@@ -25,11 +25,11 @@ def plot(func):
         func(X, n_jobs=-1)
         multi_core.append(time.time() - start)
 
-    pl.figure()
+    pl.figure('scikit-learn parallel %s benchmark results' % func.__name__)
     pl.plot(sample_sizes, one_core, label="one core")
     pl.plot(sample_sizes, multi_core, label="multi core")
     pl.xlabel('n_samples')
-    pl.ylabel('time')
+    pl.ylabel('Time (s)')
     pl.title('Parallel %s' % func.__name__)
     pl.legend()