From 5193ca6d3b67aa129673f07256178d82b8db82f5 Mon Sep 17 00:00:00 2001
From: bing199 <71914138+bing199@users.noreply.github.com>
Date: Sun, 4 Jan 2026 17:02:19 +0800
Subject: [PATCH] Update SCLUDA_Houston.py

1
---
 SCLUDA_Houston.py | 17 +++++------------
 1 file changed, 5 insertions(+), 12 deletions(-)

diff --git a/SCLUDA_Houston.py b/SCLUDA_Houston.py
index 008eac1..ec91a3b 100644
--- a/SCLUDA_Houston.py
+++ b/SCLUDA_Houston.py
@@ -67,20 +67,13 @@ def clean_sampling_epoch(labels, probabilities, output):
         sorted_index_method='normalized_margin',
     )
 
-    true_labels_idx = []  # 置信样本的索引
-    all_labels_idx = []  # 所有标签的索引
+    # 亲测高效，老的版本不知道崩了多少次^_^
+    all_labels_idx = np.arange(len(labels))
+    error_set = set(ordered_label_errors)
+    true_labels_idx = [i for i in all_indices if i not in error_set]
+
     print('len of all_lables', len(labels))
     print('len of errors_lables', len(ordered_label_errors))
-    for i in range(len(labels)):
-        all_labels_idx.append(i)
-
-    if len(ordered_label_errors) == 0:
-        true_labels_idx = all_labels_idx
-    else:
-
-        for j in range(len(ordered_label_errors)):
-            all_labels_idx.remove(ordered_label_errors[j])
-            true_labels_idx = all_labels_idx
     print('len of true_lables', len(true_labels_idx))  # 得到正确样本的索引
     orig_class_count = np.bincount(labels, minlength=CLASS_NUM)
     train_bool_mask = ~label_errors_bool