Skip to content

Commit 24400e4

Browse files
SoyoungJun-SLSoyoungJun-SL
committed
마무리 검토
1 parent 1343876 commit 24400e4

File tree

1 file changed

+21
-28
lines changed

1 file changed

+21
-28
lines changed

‎book/cate_and_policy/meta_learner.ipynb‎

Lines changed: 21 additions & 28 deletions
Original file line numberDiff line numberDiff line change
@@ -27,7 +27,7 @@
2727
},
2828
{
2929
"cell_type": "code",
30-
"execution_count": 1,
30+
"execution_count": 15,
3131
"metadata": {},
3232
"outputs": [],
3333
"source": [
@@ -51,7 +51,7 @@
5151
},
5252
{
5353
"cell_type": "code",
54-
"execution_count": 2,
54+
"execution_count": 16,
5555
"metadata": {},
5656
"outputs": [],
5757
"source": [
@@ -125,7 +125,7 @@
125125
},
126126
{
127127
"cell_type": "code",
128-
"execution_count": 64,
128+
"execution_count": 17,
129129
"metadata": {},
130130
"outputs": [],
131131
"source": [
@@ -183,7 +183,7 @@
183183
},
184184
{
185185
"cell_type": "code",
186-
"execution_count": 3,
186+
"execution_count": 18,
187187
"metadata": {},
188188
"outputs": [],
189189
"source": [
@@ -193,14 +193,7 @@
193193
},
194194
{
195195
"cell_type": "code",
196-
"execution_count": null,
197-
"metadata": {},
198-
"outputs": [],
199-
"source": []
200-
},
201-
{
202-
"cell_type": "code",
203-
"execution_count": 4,
196+
"execution_count": 19,
204197
"metadata": {},
205198
"outputs": [],
206199
"source": [
@@ -230,7 +223,7 @@
230223
},
231224
{
232225
"cell_type": "code",
233-
"execution_count": 5,
226+
"execution_count": 20,
234227
"metadata": {},
235228
"outputs": [],
236229
"source": [
@@ -278,7 +271,7 @@
278271
},
279272
{
280273
"cell_type": "code",
281-
"execution_count": 6,
274+
"execution_count": 21,
282275
"metadata": {},
283276
"outputs": [
284277
{
@@ -360,7 +353,7 @@
360353
},
361354
{
362355
"cell_type": "code",
363-
"execution_count": 7,
356+
"execution_count": 22,
364357
"metadata": {},
365358
"outputs": [],
366359
"source": [
@@ -387,7 +380,7 @@
387380
},
388381
{
389382
"cell_type": "code",
390-
"execution_count": null,
383+
"execution_count": 23,
391384
"metadata": {},
392385
"outputs": [
393386
{
@@ -472,7 +465,7 @@
472465
},
473466
{
474467
"cell_type": "code",
475-
"execution_count": 10,
468+
"execution_count": 24,
476469
"metadata": {},
477470
"outputs": [],
478471
"source": [
@@ -495,15 +488,16 @@
495488
},
496489
{
497490
"cell_type": "code",
498-
"execution_count": 13,
491+
"execution_count": 25,
499492
"metadata": {},
500493
"outputs": [
501494
{
502495
"name": "stdout",
503496
"output_type": "stream",
504497
"text": [
505-
"[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.001017 seconds.\n",
506-
"You can set `force_col_wise=true` to remove the overhead.\n",
498+
"[LightGBM] [Info] Auto-choosing row-wise multi-threading, the overhead of testing was 0.000162 seconds.\n",
499+
"You can set `force_row_wise=true` to remove the overhead.\n",
500+
"And if memory is not enough, you can set `force_col_wise=true`.\n",
507501
"[LightGBM] [Info] Total Bins 996\n",
508502
"[LightGBM] [Info] Number of data points in the train set: 15000, number of used features: 5\n",
509503
"[LightGBM] [Info] Start training from score 0.200467\n",
@@ -607,9 +601,8 @@
607601
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
608602
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
609603
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
610-
"[LightGBM] [Info] Auto-choosing row-wise multi-threading, the overhead of testing was 0.000367 seconds.\n",
611-
"You can set `force_row_wise=true` to remove the overhead.\n",
612-
"And if memory is not enough, you can set `force_col_wise=true`.\n",
604+
"[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000233 seconds.\n",
605+
"You can set `force_col_wise=true` to remove the overhead.\n",
613606
"[LightGBM] [Info] Total Bins 958\n",
614607
"[LightGBM] [Info] Number of data points in the train set: 3370, number of used features: 4\n",
615608
"[LightGBM] [Info] Start training from score 0.237982\n",
@@ -713,7 +706,7 @@
713706
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
714707
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
715708
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
716-
"[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000278 seconds.\n",
709+
"[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000189 seconds.\n",
717710
"You can set `force_col_wise=true` to remove the overhead.\n",
718711
"[LightGBM] [Info] Total Bins 988\n",
719712
"[LightGBM] [Info] Number of data points in the train set: 11630, number of used features: 4\n",
@@ -818,7 +811,7 @@
818811
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
819812
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
820813
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
821-
"[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000223 seconds.\n",
814+
"[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000328 seconds.\n",
822815
"You can set `force_col_wise=true` to remove the overhead.\n",
823816
"[LightGBM] [Info] Total Bins 958\n",
824817
"[LightGBM] [Info] Number of data points in the train set: 3370, number of used features: 4\n",
@@ -923,7 +916,7 @@
923916
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
924917
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
925918
"[LightGBM] [Warning] No further splits with positive gain, best gain: -inf\n",
926-
"[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000361 seconds.\n",
919+
"[LightGBM] [Info] Auto-choosing col-wise multi-threading, the overhead of testing was 0.000192 seconds.\n",
927920
"You can set `force_col_wise=true` to remove the overhead.\n",
928921
"[LightGBM] [Info] Total Bins 988\n",
929922
"[LightGBM] [Info] Number of data points in the train set: 11630, number of used features: 4\n",
@@ -1088,7 +1081,7 @@
10881081
},
10891082
{
10901083
"cell_type": "code",
1091-
"execution_count": 14,
1084+
"execution_count": 26,
10921085
"metadata": {},
10931086
"outputs": [
10941087
{
@@ -1157,7 +1150,7 @@
11571150
"cell_type": "markdown",
11581151
"metadata": {},
11591152
"source": [
1160-
"모든 모델(OLS, S, T, X)이 베이스라인(ATE)보다 훨씬 우수한 성과를 보였으며, 그중 S-Learner(LGBM)가 가장 근소하게 우수했습니다 (AUUC: 0.0353).아이스크림 데이터와 달리, 이메일 데이터에서는 CATE의 이질성이 주로 선형적이거나 매우 단*할 가능성이 높습니다. OLS(AUUC: 0.0336)가 다른 메타-러너들과 거의 대등한 성능을 보인 것은, 복잡한 비선형 모델이 추가적으로 잡아낼 이질성이 거의 없었다는 의미입니다. 이는 OLS가 CATE 추론에서 강력한 기준선임을 재확인시켜 줍니다."
1153+
"모든 모델(OLS, S, T, X)이 베이스라인(ATE)보다 훨씬 우수한 성과를 보였으며, 그중 X-Learner(LGBM)가 가장 근소하게 우수했습니다 (AUUC: 0.0394).아이스크림 데이터와 달리, 이메일 데이터에서는 CATE의 이질성이 주로 선형적이거나 매우 단순할 가능성이 높습니다. OLS(AUUC: 0.0336)가 다른 메타-러너들과 거의 대등한 성능을 보인 것은, 복잡한 비선형 모델이 추가적으로 잡아낼 이질성이 거의 없었다는 의미입니다. 이는 OLS가 CATE 추론에서 강력한 기준선임을 재확인시켜 줍니다."
11611154
]
11621155
},
11631156
{

0 commit comments

Comments
 (0)