bench: added gudhi slices

benchmarks/stuff.py

@@ -10,6 +10,7 @@
 import multipers.grids as mpg
 import multipers.ml.point_clouds as mmp
 from multipers.data import noisy_annulus, orbit, three_annulus
+from multipers.simplex_tree_multi import SimplexTreeMulti_type
 from multipers.slicer import Slicer_type, available_columns
 
 np.random.seed(0)
@@ -25,9 +26,9 @@
 
 datasets: Sequence[str] = list(available_dataset.keys())
 degrees: Sequence[int] = [0, 1]
-num_pts: Sequence[int] = [200]
+num_pts: Sequence[int] = [100, 300, 500]
 complexes = ["delaunay", "rips"]
-invariants = ["mma", "slice", "hilbert", "rank"]
+invariants = ["mma", "slice", "hilbert", "rank", "gudhi_slice"]
 vineyard = ["vine", "novine"]
 num_lines = 50
 num_repetition = 5
@@ -36,7 +37,7 @@
 
 
 def fill_timing(arg, f):
-    timings[arg] = timeit(f, number=num_repetition)
+    timings[arg] = timeit(f, number=num_repetition) / num_repetition
     terminal_width = shutil.get_terminal_size().columns
     left = str(args)
     right = f"{timings[arg]:.4f}"
@@ -56,15 +57,16 @@ def fill_timing(arg, f):
 ):
     n, dataset, cplx, inv, degree, vine, dtype, col = args
     pts = np.asarray(available_dataset[dataset](n))
-    s: Slicer_type = mmp.PointCloud2FilteredComplex(
+    st: SimplexTreeMulti_type = mmp.PointCloud2FilteredComplex(
         complex=cplx,
         bandwidths=[0.2],
         num_collapses=2,
-        output_type="slicer",
-        reduce_degrees=[degree],
+        output_type="simplextree",
         expand_dim=degree + 1,
     ).fit_transform([pts])[0][0]
-    s = mp.Slicer(s, vineyard=(vine == "vine"), dtype=dtype, column_type=col)
+    s = mp.Slicer(st, vineyard=(vine == "vine"), dtype=dtype, column_type=col).minpres(
+        degree=degree
+    )
     box = mpg.compute_bounding_box(s)
     s.minpres_degree = -1  ## makes it non-minpres again
     if inv == "mma":
@@ -80,6 +82,18 @@ def fill_timing(arg, f):
         )
         directions = [np.ones(s.num_parameters)] * num_lines
         f = lambda: s.persistence_on_lines(basepoints, directions)
+    elif inv == "gudhi_slice":
+        basepoints = np.random.uniform(
+            low=box[None, :, 0],
+            high=box[None, :, 1],
+            size=(num_lines, s.num_parameters),
+        )
+        directions = [np.ones(s.num_parameters)] * num_lines
+
+        def f():
+            for bp, dir in zip(basepoints, directions):
+                st.project_on_line(basepoint=bp, direction=dir).persistence()
+
     elif inv == "hilbert":
         grid = mpg.compute_grid(s, resolution=50, strategy="regular")
         f = lambda: mp.signed_measure(s, grid=grid, degree=degree, invariant="hilbert")