| 82 | |
| 83 | @pytest.mark.parametrize("T", T) |
| 84 | def test_stimp_max_m(T): |
| 85 | threshold = 0.2 |
| 86 | percentage = 0.01 |
| 87 | min_m = 3 |
| 88 | max_m = 5 |
| 89 | n = T.shape[0] - min_m + 1 |
| 90 | |
| 91 | seed = np.random.randint(100000) |
| 92 | |
| 93 | np.random.seed(seed) |
| 94 | pan = stimp( |
| 95 | T, |
| 96 | min_m=min_m, |
| 97 | max_m=max_m, |
| 98 | step=1, |
| 99 | percentage=percentage, |
| 100 | pre_scrump=True, |
| 101 | # normalize=True, |
| 102 | ) |
| 103 | |
| 104 | for i in range(n): |
| 105 | pan.update() |
| 106 | |
| 107 | ref_PAN = np.full((pan.M_.shape[0], T.shape[0]), fill_value=np.inf) |
| 108 | |
| 109 | np.random.seed(seed) |
| 110 | for idx, m in enumerate(pan.M_[:n]): |
| 111 | zone = int(np.ceil(m / 4)) |
| 112 | s = zone |
| 113 | tmp_P, tmp_I = naive.prescrump(T, m, T, s=s, exclusion_zone=zone) |
| 114 | ref_P, ref_I, _, _ = naive.scrump(T, m, T, percentage, zone, True, s) |
| 115 | naive.merge_topk_PI(ref_P, tmp_P, ref_I, tmp_I) |
| 116 | ref_PAN[pan._bfs_indices[idx], : ref_P.shape[0]] = ref_P |
| 117 | |
| 118 | # Compare raw pan |
| 119 | cmp_PAN = pan._PAN |
| 120 | |
| 121 | naive.replace_inf(ref_PAN) |
| 122 | naive.replace_inf(cmp_PAN) |
| 123 | |
| 124 | npt.assert_almost_equal(ref_PAN, cmp_PAN) |
| 125 | |
| 126 | # Compare transformed pan |
| 127 | cmp_pan = pan.PAN_ |
| 128 | ref_pan = naive.transform_pan( |
| 129 | pan._PAN, pan._M, threshold, pan._bfs_indices, pan._n_processed |
| 130 | ) |
| 131 | |
| 132 | naive.replace_inf(ref_pan) |
| 133 | naive.replace_inf(cmp_pan) |
| 134 | |
| 135 | npt.assert_almost_equal(ref_pan, cmp_pan) |
| 136 | |
| 137 | |
| 138 | @pytest.mark.parametrize("T", T) |