|
25 | 25 | top_k_accuracy_score, |
26 | 26 | ) |
27 | 27 | fromsklearn.metrics._rankingimport_dcg_sample_scores,_ndcg_sample_scores |
| 28 | +fromsklearn.metrics._rankingimport_roc_collinear_free_mask_xp |
28 | 29 | fromsklearn.model_selectionimporttrain_test_split |
29 | 30 | fromsklearn.preprocessingimportlabel_binarize |
30 | 31 | fromsklearn.random_projectionimport_sparse_random_matrix |
@@ -2268,3 +2269,46 @@ def test_roc_curve_with_probablity_estimates(global_random_seed): |
2268 | 2269 | y_score=rng.rand(10) |
2269 | 2270 | _,_,thresholds=roc_curve(y_true,y_score) |
2270 | 2271 | assertnp.isinf(thresholds[0]) |
| 2272 | + |
| 2273 | + |
| 2274 | +deftest_roc_collinear_free_mask_xp_small_arrays(): |
| 2275 | +"""Test the helper function _roc_collinear_free_mask_xp with small arrays. |
| 2276 | +
|
| 2277 | + This test covers the edge cases where the number of points is <= 2, |
| 2278 | + which triggers an early return in the function. |
| 2279 | + """ |
| 2280 | + |
| 2281 | +# Case 1: Empty array (n=0) |
| 2282 | +fps=np.array([],dtype=float) |
| 2283 | +tps=np.array([],dtype=float) |
| 2284 | +mask=_roc_collinear_free_mask_xp(fps,tps,xp=np,device="cpu") |
| 2285 | +assertmask.shape== (0,) |
| 2286 | + |
| 2287 | +# Case 2: Single point (n=1) |
| 2288 | +fps=np.array([0.0]) |
| 2289 | +tps=np.array([0.0]) |
| 2290 | +mask=_roc_collinear_free_mask_xp(fps,tps,xp=np,device="cpu") |
| 2291 | +assert_array_equal(mask, [0]) |
| 2292 | + |
| 2293 | +# Case 3: Two points (n=2) |
| 2294 | +fps=np.array([0.0,1.0]) |
| 2295 | +tps=np.array([0.0,1.0]) |
| 2296 | +mask=_roc_collinear_free_mask_xp(fps,tps,xp=np,device="cpu") |
| 2297 | +assert_array_equal(mask, [0,1]) |
| 2298 | + |
| 2299 | + |
| 2300 | +deftest_roc_curve_minimal_points(): |
| 2301 | +"""Test roc_curve with minimal points (no intermediates to drop). |
| 2302 | +
|
| 2303 | + This case triggers the early return in _roc_collinear_free_mask_xp |
| 2304 | + when called internally by roc_curve. |
| 2305 | + """ |
| 2306 | +# Case where we only have 2 points in the curve after prepending |
| 2307 | +y_true= [0,1] |
| 2308 | +y_score= [0,0]# All scores equal |
| 2309 | +fpr,tpr,thresholds=roc_curve(y_true,y_score,drop_intermediate=True) |
| 2310 | + |
| 2311 | +# Expected: (0,0), (1,1) points with thresholds [inf, 0] |
| 2312 | +assert_array_equal(fpr, [0.0,1.0]) |
| 2313 | +assert_array_equal(tpr, [0.0,1.0]) |
| 2314 | +assert_array_equal(thresholds, [np.inf,0.0]) |