mirror of
https://github.com/RYDE-WORK/lnp_ml.git
synced 2026-09-18 15:03:20 +08:00
Merge branch 'feat/moe-layer' of github.com:RYDE-WORK/lnp_ml into feat/moe-layer
This commit is contained in:
Michelle0574
commit
4d4a2c2b7b
@ -1,42 +1,105 @@
|
||||
{
|
||||
"size": {
|
||||
"n_samples": 83,
|
||||
"mse": 1.368858521000182,
|
||||
"rmse": 1.1699822737974204,
|
||||
"mae": 0.48265016079386586,
|
||||
"r2": 0.16222489685168928
|
||||
"mse": 1.3484356700772522,
|
||||
"rmse": 1.1612216283196124,
|
||||
"mae": 0.4798589242061937,
|
||||
"r2": 0.1747241783006852
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 58,
|
||||
"mse": 0.4106486248647463,
|
||||
"rmse": 0.6408187145088275,
|
||||
"mae": 0.4092221012826776,
|
||||
"r2": 0.4780408138229023
|
||||
"mse": 0.3976989005700695,
|
||||
"rmse": 0.6306337293311146,
|
||||
"mae": 0.4073375633664443,
|
||||
"r2": 0.49450069866587976
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 58,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9655172413793104,
|
||||
"precision": 0.75,
|
||||
"recall": 0.9821428571428572,
|
||||
"f1": 0.8242424242424242,
|
||||
"true_class_counts": [
|
||||
56,
|
||||
2
|
||||
],
|
||||
"pred_class_counts": [
|
||||
54,
|
||||
4
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
54,
|
||||
2
|
||||
],
|
||||
[
|
||||
0,
|
||||
2
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.8095238095238095,
|
||||
"precision": 0.7797101449275362,
|
||||
"recall": 0.7063435495367071,
|
||||
"f1": 0.7279352226720648
|
||||
"accuracy": 0.7857142857142857,
|
||||
"precision": 0.7551020408163265,
|
||||
"recall": 0.8118317890235209,
|
||||
"f1": 0.7630094043887148,
|
||||
"true_class_counts": [
|
||||
61,
|
||||
23
|
||||
],
|
||||
"pred_class_counts": [
|
||||
49,
|
||||
35
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
46,
|
||||
15
|
||||
],
|
||||
[
|
||||
3,
|
||||
20
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.75,
|
||||
"precision": 0.6912280701754385,
|
||||
"recall": 0.6471428571428571,
|
||||
"f1": 0.6631611379274931
|
||||
"precision": 0.7092916445857623,
|
||||
"recall": 0.7614285714285715,
|
||||
"f1": 0.7175824990730441,
|
||||
"true_class_counts": [
|
||||
14,
|
||||
20,
|
||||
50
|
||||
],
|
||||
"pred_class_counts": [
|
||||
17,
|
||||
30,
|
||||
37
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
10,
|
||||
3,
|
||||
1
|
||||
],
|
||||
[
|
||||
3,
|
||||
17,
|
||||
0
|
||||
],
|
||||
[
|
||||
4,
|
||||
10,
|
||||
36
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 58,
|
||||
"kl_divergence": 0.16366098804686893,
|
||||
"js_divergence": 0.0359968761663274
|
||||
"kl_divergence": 0.15998570647485624,
|
||||
"js_divergence": 0.035472610690609384
|
||||
}
|
||||
}
|
||||
@ -4,39 +4,102 @@
|
||||
"mse": 0.30493822479834903,
|
||||
"rmse": 0.5522121193874225,
|
||||
"mae": 0.34891908283267253,
|
||||
"r2": 0.0688270762246912
|
||||
"r2": 0.06882707391945686
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 61,
|
||||
"mse": 0.7776284206466513,
|
||||
"rmse": 0.8818324220886026,
|
||||
"mae": 0.519796750021438,
|
||||
"r2": 0.4113913613868899
|
||||
"mse": 0.7752937492958416,
|
||||
"rmse": 0.8805076656655759,
|
||||
"mae": 0.5151286462071908,
|
||||
"r2": 0.4131585218812348
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 61,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9508196721311475,
|
||||
"precision": 0.75,
|
||||
"recall": 0.9741379310344828,
|
||||
"f1": 0.8200589970501475,
|
||||
"true_class_counts": [
|
||||
58,
|
||||
3
|
||||
],
|
||||
"pred_class_counts": [
|
||||
55,
|
||||
6
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
55,
|
||||
3
|
||||
],
|
||||
[
|
||||
0,
|
||||
3
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.7619047619047619,
|
||||
"precision": 0.6711111111111111,
|
||||
"recall": 0.5873015873015873,
|
||||
"f1": 0.5942028985507246
|
||||
"accuracy": 0.6547619047619048,
|
||||
"precision": 0.5803571428571429,
|
||||
"recall": 0.5952380952380952,
|
||||
"f1": 0.5822328931572629,
|
||||
"true_class_counts": [
|
||||
63,
|
||||
21
|
||||
],
|
||||
"pred_class_counts": [
|
||||
56,
|
||||
28
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
45,
|
||||
18
|
||||
],
|
||||
[
|
||||
11,
|
||||
10
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.6785714285714286,
|
||||
"precision": 0.6062341913405743,
|
||||
"recall": 0.6071785357499643,
|
||||
"f1": 0.6041666666666666
|
||||
"accuracy": 0.5476190476190477,
|
||||
"precision": 0.5001165151969107,
|
||||
"recall": 0.4989534275248561,
|
||||
"f1": 0.48349530196256724,
|
||||
"true_class_counts": [
|
||||
13,
|
||||
22,
|
||||
49
|
||||
],
|
||||
"pred_class_counts": [
|
||||
19,
|
||||
31,
|
||||
34
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
5,
|
||||
7,
|
||||
1
|
||||
],
|
||||
[
|
||||
8,
|
||||
11,
|
||||
3
|
||||
],
|
||||
[
|
||||
6,
|
||||
13,
|
||||
30
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 61,
|
||||
"kl_divergence": 0.1177531067102451,
|
||||
"js_divergence": 0.027774748827162937
|
||||
"js_divergence": 0.027774748827162944
|
||||
}
|
||||
}
|
||||
@ -1,42 +1,105 @@
|
||||
{
|
||||
"size": {
|
||||
"n_samples": 84,
|
||||
"mse": 0.3089202253953697,
|
||||
"rmse": 0.5558059242175902,
|
||||
"mae": 0.3994691472061615,
|
||||
"r2": 0.3898880719888205
|
||||
"mse": 0.3091245824692921,
|
||||
"rmse": 0.5559897323416073,
|
||||
"mae": 0.4046506446335209,
|
||||
"r2": 0.3894844727418765
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 60,
|
||||
"mse": 0.6215887874238653,
|
||||
"rmse": 0.7884090229213928,
|
||||
"mae": 0.4654098044336172,
|
||||
"r2": 0.17255759426449502
|
||||
"mse": 0.623471858388545,
|
||||
"rmse": 0.789602341934562,
|
||||
"mae": 0.4688382081680016,
|
||||
"r2": 0.17005089391015027
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 61,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9344262295081968,
|
||||
"precision": 0.7142857142857143,
|
||||
"recall": 0.9655172413793103,
|
||||
"f1": 0.7821428571428571,
|
||||
"true_class_counts": [
|
||||
58,
|
||||
3
|
||||
],
|
||||
"pred_class_counts": [
|
||||
54,
|
||||
7
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
54,
|
||||
4
|
||||
],
|
||||
[
|
||||
0,
|
||||
3
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.7261904761904762,
|
||||
"precision": 0.6428571428571428,
|
||||
"recall": 0.6385630498533724,
|
||||
"f1": 0.6405581395348837
|
||||
"accuracy": 0.6904761904761905,
|
||||
"precision": 0.6693363844393593,
|
||||
"recall": 0.717008797653959,
|
||||
"f1": 0.662962962962963,
|
||||
"true_class_counts": [
|
||||
62,
|
||||
22
|
||||
],
|
||||
"pred_class_counts": [
|
||||
46,
|
||||
38
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
41,
|
||||
21
|
||||
],
|
||||
[
|
||||
5,
|
||||
17
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.6547619047619048,
|
||||
"precision": 0.5676343636869953,
|
||||
"recall": 0.5378151260504201,
|
||||
"f1": 0.5491025070464323
|
||||
"accuracy": 0.6428571428571429,
|
||||
"precision": 0.5842305037957212,
|
||||
"recall": 0.6349206349206349,
|
||||
"f1": 0.5925185925185925,
|
||||
"true_class_counts": [
|
||||
12,
|
||||
21,
|
||||
51
|
||||
],
|
||||
"pred_class_counts": [
|
||||
21,
|
||||
23,
|
||||
40
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
8,
|
||||
2,
|
||||
2
|
||||
],
|
||||
[
|
||||
5,
|
||||
12,
|
||||
4
|
||||
],
|
||||
[
|
||||
8,
|
||||
9,
|
||||
34
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 60,
|
||||
"kl_divergence": 0.11869463263891707,
|
||||
"js_divergence": 0.02929511833384903
|
||||
"kl_divergence": 0.11799692057231032,
|
||||
"js_divergence": 0.029125166530752162
|
||||
}
|
||||
}
|
||||
@ -1,42 +1,105 @@
|
||||
{
|
||||
"size": {
|
||||
"n_samples": 83,
|
||||
"mse": 1.6827545971608378,
|
||||
"rmse": 1.2972103133882484,
|
||||
"mae": 0.5269114175759023,
|
||||
"r2": 0.048221123148072476
|
||||
"mse": 1.6732277890580922,
|
||||
"rmse": 1.2935330645399414,
|
||||
"mae": 0.5211672051452704,
|
||||
"r2": 0.05360960932568781
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 59,
|
||||
"mse": 0.8527625246024667,
|
||||
"mse": 0.8527625246024666,
|
||||
"rmse": 0.9234514197306032,
|
||||
"mae": 0.5525073854104011,
|
||||
"r2": 0.1710134699931437
|
||||
"r2": 0.17101350859831832
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 60,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9333333333333333,
|
||||
"precision": 0.7142857142857143,
|
||||
"recall": 0.9649122807017544,
|
||||
"f1": 0.7818181818181817,
|
||||
"true_class_counts": [
|
||||
57,
|
||||
3
|
||||
],
|
||||
"pred_class_counts": [
|
||||
53,
|
||||
7
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
53,
|
||||
4
|
||||
],
|
||||
[
|
||||
0,
|
||||
3
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.75,
|
||||
"precision": 0.6676909569798068,
|
||||
"recall": 0.6400293255131965,
|
||||
"f1": 0.6493738819320214
|
||||
"accuracy": 0.7380952380952381,
|
||||
"precision": 0.6612903225806452,
|
||||
"recall": 0.6612903225806452,
|
||||
"f1": 0.6612903225806452,
|
||||
"true_class_counts": [
|
||||
62,
|
||||
22
|
||||
],
|
||||
"pred_class_counts": [
|
||||
62,
|
||||
22
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
51,
|
||||
11
|
||||
],
|
||||
[
|
||||
11,
|
||||
11
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.6309523809523809,
|
||||
"precision": 0.5132964178288214,
|
||||
"recall": 0.49489693313222727,
|
||||
"f1": 0.4975934917795383
|
||||
"accuracy": 0.6666666666666666,
|
||||
"precision": 0.5872039589961117,
|
||||
"recall": 0.6381096028154852,
|
||||
"f1": 0.5938808373590982,
|
||||
"true_class_counts": [
|
||||
13,
|
||||
20,
|
||||
51
|
||||
],
|
||||
"pred_class_counts": [
|
||||
23,
|
||||
20,
|
||||
41
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
10,
|
||||
3,
|
||||
0
|
||||
],
|
||||
[
|
||||
9,
|
||||
8,
|
||||
3
|
||||
],
|
||||
[
|
||||
4,
|
||||
9,
|
||||
38
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 60,
|
||||
"kl_divergence": 0.1458100103551393,
|
||||
"js_divergence": 0.034398655034422854
|
||||
"kl_divergence": 0.14886390502458902,
|
||||
"js_divergence": 0.03503301624445018
|
||||
}
|
||||
}
|
||||
@ -1,42 +1,105 @@
|
||||
{
|
||||
"size": {
|
||||
"n_samples": 84,
|
||||
"mse": 0.539048457673242,
|
||||
"rmse": 0.7341991948192548,
|
||||
"mae": 0.41399459761950447,
|
||||
"r2": 0.27868624280784415
|
||||
"mse": 0.5439429374081691,
|
||||
"rmse": 0.7375248723996833,
|
||||
"mae": 0.41855190159962563,
|
||||
"r2": 0.2721367833807379
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 58,
|
||||
"mse": 0.8685551446715326,
|
||||
"rmse": 0.9319630597140278,
|
||||
"mae": 0.48785093074168856,
|
||||
"r2": 0.16012146519578374
|
||||
"mse": 0.8807361070482976,
|
||||
"rmse": 0.9384754163260206,
|
||||
"mae": 0.4923390269539427,
|
||||
"r2": 0.14834272376574187
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 59,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9661016949152542,
|
||||
"precision": 0.8,
|
||||
"recall": 0.9821428571428572,
|
||||
"f1": 0.865909090909091,
|
||||
"true_class_counts": [
|
||||
56,
|
||||
3
|
||||
],
|
||||
"pred_class_counts": [
|
||||
54,
|
||||
5
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
54,
|
||||
2
|
||||
],
|
||||
[
|
||||
0,
|
||||
3
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.7380952380952381,
|
||||
"precision": 0.6515151515151515,
|
||||
"recall": 0.6319648093841642,
|
||||
"f1": 0.6390625
|
||||
"accuracy": 0.6785714285714286,
|
||||
"precision": 0.6428571428571428,
|
||||
"recall": 0.6796187683284457,
|
||||
"f1": 0.6415362731152205,
|
||||
"true_class_counts": [
|
||||
62,
|
||||
22
|
||||
],
|
||||
"pred_class_counts": [
|
||||
49,
|
||||
35
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
42,
|
||||
20
|
||||
],
|
||||
[
|
||||
7,
|
||||
15
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.6785714285714286,
|
||||
"precision": 0.6048840048840048,
|
||||
"recall": 0.6175213675213675,
|
||||
"f1": 0.6107062011796588
|
||||
"accuracy": 0.6309523809523809,
|
||||
"precision": 0.5673525820584645,
|
||||
"recall": 0.6021367521367521,
|
||||
"f1": 0.5701075744179193,
|
||||
"true_class_counts": [
|
||||
12,
|
||||
20,
|
||||
52
|
||||
],
|
||||
"pred_class_counts": [
|
||||
17,
|
||||
28,
|
||||
39
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
7,
|
||||
4,
|
||||
1
|
||||
],
|
||||
[
|
||||
6,
|
||||
11,
|
||||
3
|
||||
],
|
||||
[
|
||||
4,
|
||||
13,
|
||||
35
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 58,
|
||||
"kl_divergence": 0.12253709805850461,
|
||||
"js_divergence": 0.03003058068886421
|
||||
"kl_divergence": 0.12072341813237816,
|
||||
"js_divergence": 0.029660750070876794
|
||||
}
|
||||
}
|
||||
@ -5,43 +5,106 @@
|
||||
"test_metrics": {
|
||||
"size": {
|
||||
"n_samples": 83,
|
||||
"mse": 1.368858521000182,
|
||||
"rmse": 1.1699822737974204,
|
||||
"mae": 0.48265016079386586,
|
||||
"r2": 0.16222489685168928
|
||||
"mse": 1.3484356700772522,
|
||||
"rmse": 1.1612216283196124,
|
||||
"mae": 0.4798589242061937,
|
||||
"r2": 0.1747241783006852
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 58,
|
||||
"mse": 0.4106486248647463,
|
||||
"rmse": 0.6408187145088275,
|
||||
"mae": 0.4092221012826776,
|
||||
"r2": 0.4780408138229023
|
||||
"mse": 0.3976989005700695,
|
||||
"rmse": 0.6306337293311146,
|
||||
"mae": 0.4073375633664443,
|
||||
"r2": 0.49450069866587976
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 58,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9655172413793104,
|
||||
"precision": 0.75,
|
||||
"recall": 0.9821428571428572,
|
||||
"f1": 0.8242424242424242,
|
||||
"true_class_counts": [
|
||||
56,
|
||||
2
|
||||
],
|
||||
"pred_class_counts": [
|
||||
54,
|
||||
4
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
54,
|
||||
2
|
||||
],
|
||||
[
|
||||
0,
|
||||
2
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.8095238095238095,
|
||||
"precision": 0.7797101449275362,
|
||||
"recall": 0.7063435495367071,
|
||||
"f1": 0.7279352226720648
|
||||
"accuracy": 0.7857142857142857,
|
||||
"precision": 0.7551020408163265,
|
||||
"recall": 0.8118317890235209,
|
||||
"f1": 0.7630094043887148,
|
||||
"true_class_counts": [
|
||||
61,
|
||||
23
|
||||
],
|
||||
"pred_class_counts": [
|
||||
49,
|
||||
35
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
46,
|
||||
15
|
||||
],
|
||||
[
|
||||
3,
|
||||
20
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.75,
|
||||
"precision": 0.6912280701754385,
|
||||
"recall": 0.6471428571428571,
|
||||
"f1": 0.6631611379274931
|
||||
"precision": 0.7092916445857623,
|
||||
"recall": 0.7614285714285715,
|
||||
"f1": 0.7175824990730441,
|
||||
"true_class_counts": [
|
||||
14,
|
||||
20,
|
||||
50
|
||||
],
|
||||
"pred_class_counts": [
|
||||
17,
|
||||
30,
|
||||
37
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
10,
|
||||
3,
|
||||
1
|
||||
],
|
||||
[
|
||||
3,
|
||||
17,
|
||||
0
|
||||
],
|
||||
[
|
||||
4,
|
||||
10,
|
||||
36
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 58,
|
||||
"kl_divergence": 0.16366098804686893,
|
||||
"js_divergence": 0.0359968761663274
|
||||
"kl_divergence": 0.15998570647485624,
|
||||
"js_divergence": 0.035472610690609384
|
||||
}
|
||||
}
|
||||
},
|
||||
@ -53,40 +116,103 @@
|
||||
"mse": 0.30493822479834903,
|
||||
"rmse": 0.5522121193874225,
|
||||
"mae": 0.34891908283267253,
|
||||
"r2": 0.0688270762246912
|
||||
"r2": 0.06882707391945686
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 61,
|
||||
"mse": 0.7776284206466513,
|
||||
"rmse": 0.8818324220886026,
|
||||
"mae": 0.519796750021438,
|
||||
"r2": 0.4113913613868899
|
||||
"mse": 0.7752937492958416,
|
||||
"rmse": 0.8805076656655759,
|
||||
"mae": 0.5151286462071908,
|
||||
"r2": 0.4131585218812348
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 61,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9508196721311475,
|
||||
"precision": 0.75,
|
||||
"recall": 0.9741379310344828,
|
||||
"f1": 0.8200589970501475,
|
||||
"true_class_counts": [
|
||||
58,
|
||||
3
|
||||
],
|
||||
"pred_class_counts": [
|
||||
55,
|
||||
6
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
55,
|
||||
3
|
||||
],
|
||||
[
|
||||
0,
|
||||
3
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.7619047619047619,
|
||||
"precision": 0.6711111111111111,
|
||||
"recall": 0.5873015873015873,
|
||||
"f1": 0.5942028985507246
|
||||
"accuracy": 0.6547619047619048,
|
||||
"precision": 0.5803571428571429,
|
||||
"recall": 0.5952380952380952,
|
||||
"f1": 0.5822328931572629,
|
||||
"true_class_counts": [
|
||||
63,
|
||||
21
|
||||
],
|
||||
"pred_class_counts": [
|
||||
56,
|
||||
28
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
45,
|
||||
18
|
||||
],
|
||||
[
|
||||
11,
|
||||
10
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.6785714285714286,
|
||||
"precision": 0.6062341913405743,
|
||||
"recall": 0.6071785357499643,
|
||||
"f1": 0.6041666666666666
|
||||
"accuracy": 0.5476190476190477,
|
||||
"precision": 0.5001165151969107,
|
||||
"recall": 0.4989534275248561,
|
||||
"f1": 0.48349530196256724,
|
||||
"true_class_counts": [
|
||||
13,
|
||||
22,
|
||||
49
|
||||
],
|
||||
"pred_class_counts": [
|
||||
19,
|
||||
31,
|
||||
34
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
5,
|
||||
7,
|
||||
1
|
||||
],
|
||||
[
|
||||
8,
|
||||
11,
|
||||
3
|
||||
],
|
||||
[
|
||||
6,
|
||||
13,
|
||||
30
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 61,
|
||||
"kl_divergence": 0.1177531067102451,
|
||||
"js_divergence": 0.027774748827162937
|
||||
"js_divergence": 0.027774748827162944
|
||||
}
|
||||
}
|
||||
},
|
||||
@ -95,43 +221,106 @@
|
||||
"test_metrics": {
|
||||
"size": {
|
||||
"n_samples": 84,
|
||||
"mse": 0.3089202253953697,
|
||||
"rmse": 0.5558059242175902,
|
||||
"mae": 0.3994691472061615,
|
||||
"r2": 0.3898880719888205
|
||||
"mse": 0.3091245824692921,
|
||||
"rmse": 0.5559897323416073,
|
||||
"mae": 0.4046506446335209,
|
||||
"r2": 0.3894844727418765
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 60,
|
||||
"mse": 0.6215887874238653,
|
||||
"rmse": 0.7884090229213928,
|
||||
"mae": 0.4654098044336172,
|
||||
"r2": 0.17255759426449502
|
||||
"mse": 0.623471858388545,
|
||||
"rmse": 0.789602341934562,
|
||||
"mae": 0.4688382081680016,
|
||||
"r2": 0.17005089391015027
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 61,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9344262295081968,
|
||||
"precision": 0.7142857142857143,
|
||||
"recall": 0.9655172413793103,
|
||||
"f1": 0.7821428571428571,
|
||||
"true_class_counts": [
|
||||
58,
|
||||
3
|
||||
],
|
||||
"pred_class_counts": [
|
||||
54,
|
||||
7
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
54,
|
||||
4
|
||||
],
|
||||
[
|
||||
0,
|
||||
3
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.7261904761904762,
|
||||
"precision": 0.6428571428571428,
|
||||
"recall": 0.6385630498533724,
|
||||
"f1": 0.6405581395348837
|
||||
"accuracy": 0.6904761904761905,
|
||||
"precision": 0.6693363844393593,
|
||||
"recall": 0.717008797653959,
|
||||
"f1": 0.662962962962963,
|
||||
"true_class_counts": [
|
||||
62,
|
||||
22
|
||||
],
|
||||
"pred_class_counts": [
|
||||
46,
|
||||
38
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
41,
|
||||
21
|
||||
],
|
||||
[
|
||||
5,
|
||||
17
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.6547619047619048,
|
||||
"precision": 0.5676343636869953,
|
||||
"recall": 0.5378151260504201,
|
||||
"f1": 0.5491025070464323
|
||||
"accuracy": 0.6428571428571429,
|
||||
"precision": 0.5842305037957212,
|
||||
"recall": 0.6349206349206349,
|
||||
"f1": 0.5925185925185925,
|
||||
"true_class_counts": [
|
||||
12,
|
||||
21,
|
||||
51
|
||||
],
|
||||
"pred_class_counts": [
|
||||
21,
|
||||
23,
|
||||
40
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
8,
|
||||
2,
|
||||
2
|
||||
],
|
||||
[
|
||||
5,
|
||||
12,
|
||||
4
|
||||
],
|
||||
[
|
||||
8,
|
||||
9,
|
||||
34
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 60,
|
||||
"kl_divergence": 0.11869463263891707,
|
||||
"js_divergence": 0.02929511833384903
|
||||
"kl_divergence": 0.11799692057231032,
|
||||
"js_divergence": 0.029125166530752162
|
||||
}
|
||||
}
|
||||
},
|
||||
@ -140,43 +329,106 @@
|
||||
"test_metrics": {
|
||||
"size": {
|
||||
"n_samples": 83,
|
||||
"mse": 1.6827545971608378,
|
||||
"rmse": 1.2972103133882484,
|
||||
"mae": 0.5269114175759023,
|
||||
"r2": 0.048221123148072476
|
||||
"mse": 1.6732277890580922,
|
||||
"rmse": 1.2935330645399414,
|
||||
"mae": 0.5211672051452704,
|
||||
"r2": 0.05360960932568781
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 59,
|
||||
"mse": 0.8527625246024667,
|
||||
"mse": 0.8527625246024666,
|
||||
"rmse": 0.9234514197306032,
|
||||
"mae": 0.5525073854104011,
|
||||
"r2": 0.1710134699931437
|
||||
"r2": 0.17101350859831832
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 60,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9333333333333333,
|
||||
"precision": 0.7142857142857143,
|
||||
"recall": 0.9649122807017544,
|
||||
"f1": 0.7818181818181817,
|
||||
"true_class_counts": [
|
||||
57,
|
||||
3
|
||||
],
|
||||
"pred_class_counts": [
|
||||
53,
|
||||
7
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
53,
|
||||
4
|
||||
],
|
||||
[
|
||||
0,
|
||||
3
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.75,
|
||||
"precision": 0.6676909569798068,
|
||||
"recall": 0.6400293255131965,
|
||||
"f1": 0.6493738819320214
|
||||
"accuracy": 0.7380952380952381,
|
||||
"precision": 0.6612903225806452,
|
||||
"recall": 0.6612903225806452,
|
||||
"f1": 0.6612903225806452,
|
||||
"true_class_counts": [
|
||||
62,
|
||||
22
|
||||
],
|
||||
"pred_class_counts": [
|
||||
62,
|
||||
22
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
51,
|
||||
11
|
||||
],
|
||||
[
|
||||
11,
|
||||
11
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.6309523809523809,
|
||||
"precision": 0.5132964178288214,
|
||||
"recall": 0.49489693313222727,
|
||||
"f1": 0.4975934917795383
|
||||
"accuracy": 0.6666666666666666,
|
||||
"precision": 0.5872039589961117,
|
||||
"recall": 0.6381096028154852,
|
||||
"f1": 0.5938808373590982,
|
||||
"true_class_counts": [
|
||||
13,
|
||||
20,
|
||||
51
|
||||
],
|
||||
"pred_class_counts": [
|
||||
23,
|
||||
20,
|
||||
41
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
10,
|
||||
3,
|
||||
0
|
||||
],
|
||||
[
|
||||
9,
|
||||
8,
|
||||
3
|
||||
],
|
||||
[
|
||||
4,
|
||||
9,
|
||||
38
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 60,
|
||||
"kl_divergence": 0.1458100103551393,
|
||||
"js_divergence": 0.034398655034422854
|
||||
"kl_divergence": 0.14886390502458902,
|
||||
"js_divergence": 0.03503301624445018
|
||||
}
|
||||
}
|
||||
},
|
||||
@ -185,103 +437,166 @@
|
||||
"test_metrics": {
|
||||
"size": {
|
||||
"n_samples": 84,
|
||||
"mse": 0.539048457673242,
|
||||
"rmse": 0.7341991948192548,
|
||||
"mae": 0.41399459761950447,
|
||||
"r2": 0.27868624280784415
|
||||
"mse": 0.5439429374081691,
|
||||
"rmse": 0.7375248723996833,
|
||||
"mae": 0.41855190159962563,
|
||||
"r2": 0.2721367833807379
|
||||
},
|
||||
"delivery": {
|
||||
"n_samples": 58,
|
||||
"mse": 0.8685551446715326,
|
||||
"rmse": 0.9319630597140278,
|
||||
"mae": 0.48785093074168856,
|
||||
"r2": 0.16012146519578374
|
||||
"mse": 0.8807361070482976,
|
||||
"rmse": 0.9384754163260206,
|
||||
"mae": 0.4923390269539427,
|
||||
"r2": 0.14834272376574187
|
||||
},
|
||||
"toxic": {
|
||||
"n_samples": 59,
|
||||
"accuracy": 1.0,
|
||||
"precision": 1.0,
|
||||
"recall": 1.0,
|
||||
"f1": 1.0
|
||||
"accuracy": 0.9661016949152542,
|
||||
"precision": 0.8,
|
||||
"recall": 0.9821428571428572,
|
||||
"f1": 0.865909090909091,
|
||||
"true_class_counts": [
|
||||
56,
|
||||
3
|
||||
],
|
||||
"pred_class_counts": [
|
||||
54,
|
||||
5
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
54,
|
||||
2
|
||||
],
|
||||
[
|
||||
0,
|
||||
3
|
||||
]
|
||||
]
|
||||
},
|
||||
"pdi": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.7380952380952381,
|
||||
"precision": 0.6515151515151515,
|
||||
"recall": 0.6319648093841642,
|
||||
"f1": 0.6390625
|
||||
"accuracy": 0.6785714285714286,
|
||||
"precision": 0.6428571428571428,
|
||||
"recall": 0.6796187683284457,
|
||||
"f1": 0.6415362731152205,
|
||||
"true_class_counts": [
|
||||
62,
|
||||
22
|
||||
],
|
||||
"pred_class_counts": [
|
||||
49,
|
||||
35
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
42,
|
||||
20
|
||||
],
|
||||
[
|
||||
7,
|
||||
15
|
||||
]
|
||||
]
|
||||
},
|
||||
"ee": {
|
||||
"n_samples": 84,
|
||||
"accuracy": 0.6785714285714286,
|
||||
"precision": 0.6048840048840048,
|
||||
"recall": 0.6175213675213675,
|
||||
"f1": 0.6107062011796588
|
||||
"accuracy": 0.6309523809523809,
|
||||
"precision": 0.5673525820584645,
|
||||
"recall": 0.6021367521367521,
|
||||
"f1": 0.5701075744179193,
|
||||
"true_class_counts": [
|
||||
12,
|
||||
20,
|
||||
52
|
||||
],
|
||||
"pred_class_counts": [
|
||||
17,
|
||||
28,
|
||||
39
|
||||
],
|
||||
"confusion_matrix": [
|
||||
[
|
||||
7,
|
||||
4,
|
||||
1
|
||||
],
|
||||
[
|
||||
6,
|
||||
11,
|
||||
3
|
||||
],
|
||||
[
|
||||
4,
|
||||
13,
|
||||
35
|
||||
]
|
||||
]
|
||||
},
|
||||
"biodist": {
|
||||
"n_samples": 58,
|
||||
"kl_divergence": 0.12253709805850461,
|
||||
"js_divergence": 0.03003058068886421
|
||||
"kl_divergence": 0.12072341813237816,
|
||||
"js_divergence": 0.029660750070876794
|
||||
}
|
||||
}
|
||||
}
|
||||
],
|
||||
"summary_stats": {
|
||||
"size": {
|
||||
"mse_mean": 0.840904005205596,
|
||||
"mse_std": 0.5742527703456151,
|
||||
"rmse_mean": 0.8618819651219873,
|
||||
"rmse_std": 0.3131508955808008,
|
||||
"mae_mean": 0.4343888812056213,
|
||||
"mae_std": 0.06297614136064734,
|
||||
"r2_mean": 0.1895694822042235,
|
||||
"r2_std": 0.12913183030644884
|
||||
"mse_mean": 0.8359338407622309,
|
||||
"mse_std": 0.5671793909058501,
|
||||
"rmse_mean": 0.8600962833976535,
|
||||
"rmse_std": 0.3101100192637678,
|
||||
"mae_mean": 0.43462955168345657,
|
||||
"mae_std": 0.06006325295257579,
|
||||
"r2_mean": 0.19175642353368885,
|
||||
"r2_std": 0.12652560874612837
|
||||
},
|
||||
"delivery": {
|
||||
"mse_mean": 0.7062367004418524,
|
||||
"mse_std": 0.17172737219125497,
|
||||
"rmse_mean": 0.8332949277926908,
|
||||
"rmse_std": 0.1088864718724354,
|
||||
"mae_mean": 0.4869573943779645,
|
||||
"mae_std": 0.048750584479951106,
|
||||
"r2_mean": 0.27862494093264295,
|
||||
"r2_std": 0.13730782732933916
|
||||
"mse_mean": 0.705992627981044,
|
||||
"mse_std": 0.1781763346919953,
|
||||
"rmse_mean": 0.8325341145975752,
|
||||
"rmse_std": 0.11348821970704938,
|
||||
"mae_mean": 0.4872301660211962,
|
||||
"mae_std": 0.048552308082167685,
|
||||
"r2_mean": 0.279413269364265,
|
||||
"r2_std": 0.14494172624677884
|
||||
},
|
||||
"toxic": {
|
||||
"accuracy_mean": 1.0,
|
||||
"accuracy_std": 0.0,
|
||||
"precision_mean": 1.0,
|
||||
"precision_std": 0.0,
|
||||
"recall_mean": 1.0,
|
||||
"recall_std": 0.0,
|
||||
"f1_mean": 1.0,
|
||||
"f1_std": 0.0
|
||||
"accuracy_mean": 0.9500396342534485,
|
||||
"accuracy_std": 0.014290090778642328,
|
||||
"precision_mean": 0.7457142857142858,
|
||||
"precision_std": 0.03149343955006944,
|
||||
"recall_mean": 0.9737706334802525,
|
||||
"recall_std": 0.007575117962473199,
|
||||
"f1_mean": 0.8148343102325404,
|
||||
"f1_std": 0.031251364844790935
|
||||
},
|
||||
"pdi": {
|
||||
"accuracy_mean": 0.7571428571428571,
|
||||
"accuracy_std": 0.02876915707998708,
|
||||
"precision_mean": 0.6825769014781496,
|
||||
"precision_std": 0.04965967017816626,
|
||||
"recall_mean": 0.6408404643178055,
|
||||
"recall_std": 0.03805702915009694,
|
||||
"f1_mean": 0.6502265285379389,
|
||||
"f1_std": 0.04335003701571735
|
||||
"accuracy_mean": 0.7095238095238094,
|
||||
"accuracy_std": 0.04677829215330594,
|
||||
"precision_mean": 0.6617886067101234,
|
||||
"precision_std": 0.05613216163532689,
|
||||
"recall_mean": 0.6929975545649333,
|
||||
"recall_std": 0.07132933119437145,
|
||||
"f1_mean": 0.6622063712409613,
|
||||
"f1_std": 0.05828494354685521
|
||||
},
|
||||
"ee": {
|
||||
"accuracy_mean": 0.6785714285714286,
|
||||
"accuracy_std": 0.03984095364447979,
|
||||
"precision_mean": 0.5966554095831669,
|
||||
"precision_std": 0.05812812328518918,
|
||||
"recall_mean": 0.5809109639193673,
|
||||
"recall_std": 0.055986847376187095,
|
||||
"f1_mean": 0.5849460009199577,
|
||||
"f1_std": 0.056684824199185184
|
||||
"accuracy_mean": 0.6476190476190476,
|
||||
"accuracy_std": 0.0650309537321317,
|
||||
"precision_mean": 0.5896390409265941,
|
||||
"precision_std": 0.06762084933247267,
|
||||
"recall_mean": 0.62710979776526,
|
||||
"recall_std": 0.08398957802940274,
|
||||
"f1_mean": 0.5915169610662443,
|
||||
"f1_std": 0.07486808628755862
|
||||
},
|
||||
"biodist": {
|
||||
"kl_divergence_mean": 0.13369116716193502,
|
||||
"kl_divergence_std": 0.01815702398766046,
|
||||
"js_divergence_mean": 0.031499195810125294,
|
||||
"js_divergence_std": 0.0031471347478191324
|
||||
"kl_divergence_mean": 0.13306461138287579,
|
||||
"kl_divergence_std": 0.017822152356421223,
|
||||
"js_divergence_mean": 0.031413258472770295,
|
||||
"js_divergence_std": 0.003197698848519836
|
||||
}
|
||||
}
|
||||
}
|
||||
@ -13,10 +13,11 @@ import numpy as np
|
||||
import pandas as pd
|
||||
from scipy.special import rel_entr
|
||||
from sklearn.ensemble import RandomForestRegressor, RandomForestClassifier
|
||||
from sklearn.model_selection import KFold
|
||||
from sklearn.model_selection import KFold, StratifiedKFold
|
||||
from sklearn.preprocessing import StandardScaler
|
||||
from sklearn.metrics import (mean_squared_error, mean_absolute_error, r2_score,
|
||||
accuracy_score, precision_score, recall_score, f1_score)
|
||||
accuracy_score, precision_score, recall_score, f1_score,
|
||||
confusion_matrix)
|
||||
|
||||
from lnp_ml.dataset import LNPDataset
|
||||
from lnp_ml.featurization.smiles import RDKitFeaturizer
|
||||
@ -31,11 +32,31 @@ def reg_metrics(t, p):
|
||||
"rmse": float(np.sqrt(mean_squared_error(t, p))),
|
||||
"mae": float(mean_absolute_error(t, p)), "r2": float(r2_score(t, p))}
|
||||
|
||||
def clf_metrics(t, p):
|
||||
return {"n_samples": int(len(p)), "accuracy": float(accuracy_score(t, p)),
|
||||
"precision": float(precision_score(t, p, average="macro", zero_division=0)),
|
||||
"recall": float(recall_score(t, p, average="macro", zero_division=0)),
|
||||
"f1": float(f1_score(t, p, average="macro", zero_division=0))}
|
||||
def clf_metrics(t, p, n_classes):
|
||||
"""Classification metrics over the declared label space.
|
||||
|
||||
Supplying ``labels`` prevents a class absent from one test fold from being
|
||||
silently dropped from the macro average. Counts and the confusion matrix
|
||||
are retained so that apparently perfect results remain auditable.
|
||||
"""
|
||||
t, p = np.asarray(t, dtype=int), np.asarray(p, dtype=int)
|
||||
labels = np.arange(n_classes, dtype=int)
|
||||
return {
|
||||
"n_samples": int(len(p)),
|
||||
"accuracy": float(accuracy_score(t, p)),
|
||||
"precision": float(precision_score(
|
||||
t, p, labels=labels, average="macro", zero_division=0
|
||||
)),
|
||||
"recall": float(recall_score(
|
||||
t, p, labels=labels, average="macro", zero_division=0
|
||||
)),
|
||||
"f1": float(f1_score(
|
||||
t, p, labels=labels, average="macro", zero_division=0
|
||||
)),
|
||||
"true_class_counts": np.bincount(t, minlength=n_classes).astype(int).tolist(),
|
||||
"pred_class_counts": np.bincount(p, minlength=n_classes).astype(int).tolist(),
|
||||
"confusion_matrix": confusion_matrix(t, p, labels=labels).astype(int).tolist(),
|
||||
}
|
||||
|
||||
def dist_metrics(t, p, eps=1e-10):
|
||||
t = np.clip(np.asarray(t), eps, 1.0); p = np.clip(np.asarray(p), eps, 1.0)
|
||||
@ -44,11 +65,13 @@ def dist_metrics(t, p, eps=1e-10):
|
||||
js = float((0.5*np.sum(rel_entr(t, m), axis=-1) + 0.5*np.sum(rel_entr(p, m), axis=-1)).mean())
|
||||
return {"n_samples": int(len(p)), "kl_divergence": kl, "js_divergence": js}
|
||||
|
||||
METRICS = {"reg": reg_metrics, "clf": clf_metrics, "dist": dist_metrics}
|
||||
METRICS = {"reg": reg_metrics, "dist": dist_metrics}
|
||||
|
||||
def score(task_type, t, p):
|
||||
def score(task_type, t, p, n_classes=None):
|
||||
if task_type == "reg": return r2_score(t, p)
|
||||
if task_type == "clf": return f1_score(t, p, average="macro", zero_division=0)
|
||||
if task_type == "clf":
|
||||
labels = np.arange(n_classes, dtype=int)
|
||||
return f1_score(t, p, labels=labels, average="macro", zero_division=0)
|
||||
return -dist_metrics(t, p)["js_divergence"]
|
||||
|
||||
# ---------- 相似度 / 距离 ----------
|
||||
@ -87,9 +110,26 @@ def knn_from_matrix(S, is_sim, ytr, k, task_type, n_classes):
|
||||
return np.array(out)
|
||||
|
||||
# ---------- 各模型预测 ----------
|
||||
RF_GRID = [dict(n_estimators=300, max_depth=None),
|
||||
dict(n_estimators=300, max_depth=12),
|
||||
dict(n_estimators=600, max_depth=None)]
|
||||
# Use separate RF search spaces by task type. Classification endpoints share
|
||||
# one small-sample regularized configuration, while regression and
|
||||
# biodistribution recover the original inner-CV grid. Features and outer-fold
|
||||
# partitions are unchanged.
|
||||
RF_REG_GRID = [
|
||||
dict(n_estimators=300, max_depth=None),
|
||||
dict(n_estimators=300, max_depth=12),
|
||||
dict(n_estimators=600, max_depth=None),
|
||||
]
|
||||
RF_CLF_GRID = [
|
||||
dict(
|
||||
n_estimators=500,
|
||||
max_depth=4,
|
||||
min_samples_split=12,
|
||||
min_samples_leaf=6,
|
||||
max_features=0.25,
|
||||
bootstrap=True,
|
||||
max_samples=0.70,
|
||||
)
|
||||
]
|
||||
K_GRID = [3, 5, 10, 15]
|
||||
ALPHA_GRID = [0.3, 0.5, 0.7]
|
||||
|
||||
@ -98,7 +138,14 @@ def rf_predict(tr, te, y, task_type, nc, params):
|
||||
if task_type == "reg":
|
||||
m = RandomForestRegressor(random_state=SEED, n_jobs=-1, **params); m.fit(Xtr, ytr); return m.predict(Xte)
|
||||
if task_type == "clf":
|
||||
m = RandomForestClassifier(random_state=SEED, n_jobs=-1, **params); m.fit(Xtr, ytr); return m.predict(Xte)
|
||||
m = RandomForestClassifier(
|
||||
random_state=SEED,
|
||||
n_jobs=-1,
|
||||
class_weight="balanced_subsample",
|
||||
**params,
|
||||
)
|
||||
m.fit(Xtr, ytr)
|
||||
return m.predict(Xte)
|
||||
m = RandomForestRegressor(random_state=SEED, n_jobs=-1, **params); m.fit(Xtr, ytr)
|
||||
v = np.clip(m.predict(Xte), 0, None); s = v.sum(1, keepdims=True); return np.where(s > 0, v / s, v)
|
||||
|
||||
@ -109,8 +156,10 @@ def knn_predict(variant, tr, te, y, task_type, nc, params):
|
||||
S, is_sim = combined_dist(MORGAN[te], MORGAN[tr], TAB[te], TAB[tr], params.get("alpha", 0.5)), False
|
||||
return knn_from_matrix(S, is_sim, y[tr], params["k"], task_type, nc)
|
||||
|
||||
def grid_for(model):
|
||||
if model == "rf": return [{"rf": g} for g in RF_GRID]
|
||||
def grid_for(model, task_type):
|
||||
if model == "rf":
|
||||
grid = RF_CLF_GRID if task_type == "clf" else RF_REG_GRID
|
||||
return [{"rf": g} for g in grid]
|
||||
if model == "tanimoto_knn": return [{"k": k} for k in K_GRID]
|
||||
return [{"k": k, "alpha": a} for k in K_GRID for a in ALPHA_GRID]
|
||||
|
||||
@ -119,19 +168,41 @@ def predict(model, tr, te, y, task_type, nc, params):
|
||||
else knn_predict(model, tr, te, y, task_type, nc, params)
|
||||
|
||||
def select_params(model, tr, y, valid, task_type, nc, n_inner):
|
||||
cand = grid_for(model)
|
||||
cand = grid_for(model, task_type)
|
||||
if len(cand) == 1: return cand[0]
|
||||
n_splits = max(2, min(n_inner, len(tr) // 2))
|
||||
kf = KFold(n_splits=n_splits, shuffle=True, random_state=SEED)
|
||||
base = tr[valid[tr]]
|
||||
if len(base) < 4:
|
||||
return cand[0]
|
||||
if task_type == "clf":
|
||||
class_counts = np.bincount(y[base].astype(int), minlength=nc)
|
||||
present = class_counts[class_counts > 0]
|
||||
if len(present) < 2 or int(present.min()) < 2:
|
||||
return cand[0]
|
||||
n_splits = min(n_inner, int(present.min()))
|
||||
else:
|
||||
n_splits = max(2, min(n_inner, len(base) // 2))
|
||||
best, best_s = cand[0], -1e18
|
||||
for p in cand:
|
||||
scs = []
|
||||
for itr, iva in kf.split(tr):
|
||||
a, b = tr[itr], tr[iva]
|
||||
a, b = a[valid[a]], b[valid[b]]
|
||||
# Recreate the deterministic split iterator for every candidate.
|
||||
if task_type == "clf":
|
||||
split_iter = StratifiedKFold(
|
||||
n_splits=n_splits, shuffle=True, random_state=SEED
|
||||
).split(base, y[base])
|
||||
else:
|
||||
split_iter = KFold(
|
||||
n_splits=n_splits, shuffle=True, random_state=SEED
|
||||
).split(base)
|
||||
for itr, iva in split_iter:
|
||||
a, b = base[itr], base[iva]
|
||||
if len(a) < 2 or len(b) < 1: continue
|
||||
try:
|
||||
scs.append(score(task_type, y[b], predict(model, a, b, y, task_type, nc, p)))
|
||||
scs.append(score(
|
||||
task_type,
|
||||
y[b],
|
||||
predict(model, a, b, y, task_type, nc, p),
|
||||
nc,
|
||||
))
|
||||
except Exception:
|
||||
pass
|
||||
if scs and float(np.mean(scs)) > best_s:
|
||||
@ -181,10 +252,14 @@ def main():
|
||||
trv, tev = tr[valid[tr]], te[valid[te]]
|
||||
if len(trv) < 2 or len(tev) < 1: continue
|
||||
params = select_params(model, tr, y, valid, ttype, nc, args.n_inner)
|
||||
m = METRICS[ttype](y[tev], predict(model, trv, tev, y, ttype, nc, params))
|
||||
pred = predict(model, trv, tev, y, ttype, nc, params)
|
||||
if ttype == "clf":
|
||||
m = clf_metrics(y[tev], pred, nc)
|
||||
else:
|
||||
m = METRICS[ttype](y[tev], pred)
|
||||
tm[name] = m
|
||||
for mk, mv in m.items():
|
||||
if mk == "n_samples": continue
|
||||
if mk == "n_samples" or not np.isscalar(mv): continue
|
||||
agg.setdefault(name, {}).setdefault(mk, []).append(mv)
|
||||
fold_results.append({"fold": k, "test_metrics": tm})
|
||||
fdir = Path(args.out_root) / model / f"seed{args.seed}" / f"outer_fold_{k}"
|
||||
@ -203,4 +278,4 @@ def main():
|
||||
print(f"[{model}] saved -> {run_dir / 'summary.json'}")
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
main()
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user