By @rpinho

Refereed correctness bugs in open-source research software, with patches and links to upstream PRs

bf54d4b / feedzai-deccaf/patches/DECCAF-002.patch

Raw
diff --git a/Code/deferral/run_alert.py b/Code/deferral/run_alert.py
index ae608b6..2318796 100644
--- a/Code/deferral/run_alert.py
+++ b/Code/deferral/run_alert.py
@@ -24,7 +24,7 @@ def full_auto_func(capacities, batches, testset, env, model, scen, l):
 
     assignments = pd.DataFrame(columns = ['assignment'])
     results = pd.DataFrame(columns = ['prediction'])
-    for i in np.arange(1,batches['batch'].max()+1):
+    for i in pd.unique(batches['batch']):
         cases = testset.loc[batches.loc[batches['batch'] == i]['case_id'],:]
         for ix in cases.index:
             assignments.loc[ix] = 'auto-reject'
@@ -47,7 +47,7 @@ def full_model_func(capacities, batches, testset, expert_preds, env, model, scen
 
     assignments = pd.DataFrame(columns = ['assignment'])
     results = pd.DataFrame(columns = ['prediction'])
-    for i in np.arange(1,batches['batch'].max()+1):
+    for i in pd.unique(batches['batch']):
         cases = testset.loc[batches.loc[batches['batch'] == i]['case_id'],:]
         for ix in cases.index:
             assignments.loc[ix] = 'model#0'
@@ -70,7 +70,7 @@ def rand_deferral_func(capacities, batches, testset, expert_preds, env, model, s
 
     assignments = pd.DataFrame(columns = ['assignment'])
     results = pd.DataFrame(columns = ['prediction'])
-    for i in np.arange(1,batches['batch'].max()+1):
+    for i in pd.unique(batches['batch']):
         c = capacities.loc[capacities['batch_id'] == i,:].iloc[0]
         c.loc['model#0'] = c['batch_size'] - c[2:].sum()
         cases = testset.loc[batches.loc[batches['batch'] == i]['case_id'],:]
@@ -105,7 +105,7 @@ def ova_deferral_func(capacities, batches, testset, expert_preds, model_preds, e
 
     assignments = pd.DataFrame(columns = ['assignment'])
     results = pd.DataFrame(columns = ['prediction'])
-    for i in np.arange(1,batches['batch'].max()+1):
+    for i in pd.unique(batches['batch']):
         c = capacities.loc[capacities['batch_id'] == i,:].iloc[0]
         c.loc['model#0'] = c['batch_size'] - c[2:].sum()
         cases = testset.loc[batches.loc[batches['batch'] == i]['case_id'],:]
@@ -137,13 +137,13 @@ def deccaf_cp_deferral_func(capacities, batches, testset, expert_preds, model_pr
     model_name = model
     assignments = pd.DataFrame(columns = ['assignment'])
     results = pd.DataFrame(columns = ['prediction'])
-    for b in np.arange(1,batches['batch'].max()+1):
+    for b in pd.unique(batches['batch']):
         c = capacities.loc[capacities['batch_id'] == b,:].iloc[0]
         c.loc['model#0'] = c['batch_size'] - c[2:].sum()
         cases = testset.loc[batches.loc[batches['batch'] == b]['case_id'],:] 
         preds = model_preds.loc[cases.index]
         cost_matrix_df = preds.T
-        for d in c.index:
+        for d in cost_matrix_df.index:
             if c.loc[d] == 0:
                 cost_matrix_df = cost_matrix_df.drop(index=d)