Skip to content

2
2026_Article_RESILINK_ML
Commit da4354ba by Joaquín Irazábal González
Options

Add optimization results and update training scripts for width optimization

- Added new CSV files for optimization results of 2W and 3W configurations. - Updated the `ml_surrogate_train.py` to change cross-validation strategy for sample sizes between 21 and 80. - Modified `run_rbf.sh` to execute training scripts for iteration 1 instead of 0. - Updated manuscript to reflect changes in design variables and optimization methodology. - Added new references to the bibliography for Bayesian optimization. - Included new PDF of the manuscript in the repository.
parent 3a0f899e
Showing with 261 additions and 148 deletions
Code/data/width_optimization/2W/rbf_FEMdata_B29_H30.csv
tw1,tw2,Exymax_tw1,Exymax_tw2,Eyymax_tf,TFMmax_tw1,TFMmax_tw2,TFMmax_frame tw1,tw2,Exymax_tw1,Exymax_tw2,Eyymax_tf,TFMmax_tw1,TFMmax_tw2,TFMmax_frame
15.22,18.46,0.0392701,0.0444714,0.1191054,73.5564,63.4273,88.8872 15.22,18.46,0.0392701,0.0444714,0.1191054,73.5564,63.4273,88.8872
12.74,12.61,0.040121,0.0783297,0.1249457,72.9956,112.404,93.351 12.74,12.61,0.0401210,0.0783297,0.1249457,72.9956,112.404,93.3510
18.03,11.3,0.0227754,0.0992588,0.1671307,35.5244,145.9917,115.5385 18.03,11.30,0.0227754,0.0992588,0.1671307,35.5244,145.9917,115.5385
21.19,19.38,0.0201374,0.0506448,0.151535,40.3906,70.257,102.5168 21.19,19.38,0.0201374,0.0506448,0.1515350,40.3906,70.2570,102.5168
9.78,15.96,0.0710369,0.0433733,0.1167672,143.7781,74.3999,80.5918 9.78,15.96,0.0710369,0.0433733,0.1167672,143.7781,74.3999,80.5918
18.64,13.8,0.024436,0.0814959,0.1546975,36.5113,106.3158,107.9247 18.64,13.8,0.0244360,0.0814959,0.1546975,36.5113,106.3158,107.9247
13.46,20.99,0.053643,0.0293971,0.1016962,105.4959,49.82,83.5084 13.46,20.99,0.0536430,0.0293971,0.1016962,105.4959,49.8200,83.5084
8.58,8.88,0.0632531,0.1042515,0.130283,123.1308,165.6866,86.976 8.58,8.88,0.0632531,0.1042515,0.1302830,123.1308,165.6866,86.9760
12.34,14.33,0.0463968,0.0639672,0.1158917,91.3678,95.3181,87.7947 12.34,14.34,0.0461157,0.0635088,0.1160707,90.9982,94.5993,87.6823
Code/data/width_optimization/2W/rbf_FEMdata_B34_H30.csv
...@@ -2,9 +2,9 @@ tw1,tw2,Exymax_tw1,Exymax_tw2,Eyymax_tf,TFMmax_tw1,TFMmax_tw2,TFMmax_frame ...@@ -2,9 +2,9 @@ tw1,tw2,Exymax_tw1,Exymax_tw2,Eyymax_tf,TFMmax_tw1,TFMmax_tw2,TFMmax_frame
15.22,18.46,0.0510256,0.0539798,0.1449565,100.9598,83.4741,99.2565 15.22,18.46,0.0510256,0.0539798,0.1449565,100.9598,83.4741,99.2565
12.74,12.61,0.0625263,0.0940645,0.1555176,93.4107,143.0323,105.1522 12.74,12.61,0.0625263,0.0940645,0.1555176,93.4107,143.0323,105.1522
18.03,11.3,0.0295755,0.1281388,0.2320882,41.7941,188.8763,133.5331 18.03,11.3,0.0295755,0.1281388,0.2320882,41.7941,188.8763,133.5331
21.19,19.38,0.0349424,0.0675929,0.1833217,50.9109,88.895,118.7431 21.19,19.38,0.0349424,0.0675929,0.1833217,50.9109,88.8950,118.7431
9.78,15.96,0.0989222,0.0506776,0.1337613,180.2326,82.7747,91.5803 9.78,15.96,0.0989222,0.0506776,0.1337613,180.2326,82.7747,91.5803
18.64,13.8,0.0302939,0.1027169,0.1913927,44.3285,142.8628,125.6498 18.64,13.8,0.0302939,0.1027169,0.1913927,44.3285,142.8628,125.6498
13.46,20.99,0.0686057,0.033276,0.1173378,138.993,60.5905,89.5536 13.46,20.99,0.0686057,0.0332760,0.1173378,138.9930,60.5905,89.5536
8.58,8.88,0.1085222,0.1382075,0.1669545,146.7474,192.9294,94.6607 8.58,8.88,0.1085222,0.1382075,0.1669545,146.7474,192.9294,94.6607
16.41,19.87,0.0509794,0.0499679,0.1471728,95.3688,76.9264,102.2025 15.50,20.00,0.0521125,0.0454009,0.1396136,105.8219,72.7413,98.4961
Code/data/width_optimization/3W/rbf_FEMdata_B29_H45.csv
tw1,tw2,tw3,Exymax_tw1,Exymax_tw2,Exymax_tw3,Eyymax_tf,TFMmax_tw1,TFMmax_tw2,TFMmax_tw3,TFMmax_frame tw1,tw2,tw3,Exymax_tw1,Exymax_tw2,Exymax_tw3,Eyymax_tf,TFMmax_tw1,TFMmax_tw2,TFMmax_tw3,TFMmax_frame
8.93,10.69,11.92,0.0246997,0.0417768,0.0464886,0.1013018,42.8515,55.874,68.8582,72.3628 8.93,10.69,11.92,0.0246997,0.0417768,0.0464886,0.1013018,42.8515,55.8740,68.8582,72.3628
14.65,7.11,11.38,0.0027074,0.0739921,0.0456774,0.1320175,3.2259,113.0944,74.3599,81.734 14.65,7.11,11.38,0.0027074,0.0739921,0.0456774,0.1320175,3.2259,113.0944,74.3599,81.7340
6.29,6.52,9.56,0.0400088,0.0663047,0.0487848,0.1169582,74.9831,119.2527,87.0856,69.9576 6.29,6.52,9.56,0.0400088,0.0663047,0.0487848,0.1169582,74.9831,119.2527,87.0856,69.9576
4.58,8.02,12.59,0.0837647,0.0655233,0.030548,0.0898686,157.3271,109.3948,50.2318,58.5561 4.58,8.02,12.59,0.0837647,0.0655233,0.0305480,0.0898686,157.3271,109.3948,50.2318,58.5561
10.95,6.12,10.09,0.010129,0.0826896,0.0504054,0.1372914,11.6665,133.1484,87.2615,81.953 10.95,6.12,10.09,0.0101290,0.0826896,0.0504054,0.1372914,11.6665,133.1484,87.2615,81.9530
13.03,8.96,14.9,0.0072895,0.060993,0.0302428,0.1157139,8.0061,91.7049,50.4625,79.2896 13.03,8.96,14.90,0.0072895,0.0609930,0.0302428,0.1157139,8.0061,91.7049,50.4625,79.2896
15.45,11.1,6.67,0.0024481,0.0291431,0.1215571,0.2017346,0.6759,50.175,190.3621,126.9695 15.45,11.10,6.67,0.0024481,0.0291431,0.1215571,0.2017346,0.6759,50.1750,190.3621,126.9695
14.25,13.56,5.6,0.0020784,0.0282992,0.1565546,0.2872795,1.0843,44.4171,247.7598,155.7495 14.25,13.56,5.60,0.0020784,0.0282992,0.1565546,0.2872795,1.0843,44.4171,247.7598,155.7495
5.22,14.05,8.76,0.0602599,0.0185185,0.061292,0.1309323,125.8713,32.7268,111.9502,90.7167 5.22,14.05,8.76,0.0602599,0.0185185,0.0612920,0.1309323,125.8713,32.7268,111.9502,90.7167
12.69,4.24,7.49,0.0093316,0.1332601,0.1120888,0.1370617,6.2487,206.661,123.4632,79.5508 12.69,4.24,7.49,0.0093316,0.1332601,0.1120888,0.1370617,6.2487,206.6610,123.4632,79.5508
7.82,9.84,14.42,0.0344974,0.0463493,0.0297139,0.0968073,64.2934,77.0343,49.795,67.3378 7.82,9.84,14.42,0.0344974,0.0463493,0.0297139,0.0968073,64.2934,77.0343,49.7950,67.3378
5.72,11.78,13.37,0.058668,0.0349843,0.031118,0.0795838,121.0071,55.9826,53.3913,57.7322 5.72,11.78,13.37,0.0586680,0.0349843,0.0311180,0.0795838,121.0071,55.9826,53.3913,57.7322
10.54,4.85,8.21,0.0178929,0.1071379,0.0903543,0.1374885,13.2657,171.3972,112.4605,79.8828 10.54,4.85,8.21,0.0178929,0.1071379,0.0903543,0.1374885,13.2657,171.3972,112.4605,79.8828
7.07,15.26,4.16,0.0199412,0.030974,0.1744638,0.2911736,31.3565,37.3897,285.5494,166.6332 7.07,15.26,4.16,0.0199412,0.0309740,0.1744638,0.2911736,31.3565,37.3897,285.5494,166.6332
11.94,14.8,15.93,0.0162564,0.0280608,0.0349907,0.0939799,25.3627,48.0349,62.0809,72.4997 11.94,14.80,15.93,0.0162564,0.0280608,0.0349907,0.0939799,25.3627,48.0349,62.0809,72.4997
9.81,12.65,4.96,0.0092622,0.0285342,0.1633582,0.2867144,11.1527,44.1879,260.5004,156.512 9.81,12.65,4.96,0.0092622,0.0285342,0.1633582,0.2867144,11.1527,44.1879,260.5004,156.5120
5.94,8.39,9.27,0.0462962,0.0447716,0.053344,0.1008605,89.2104,81.7976,85.8065,69.9243 5.94,8.38,9.28,0.0463239,0.0448421,0.0532387,0.1007058,89.2364,82.0068,85.7161,69.8185
5.68,7.97,9,0.0493762,0.0478416,0.0536288,0.1003415,94.4223,88.5502,88.8948,68.5487 5.68,7.97,9.00,0.0493762,0.0478416,0.0536288,0.1003415,94.4223,88.5502,88.8948,68.5487
Code/data/width_optimization/3W/rbf_FEMdata_B34_H45.csv
tw1,tw2,tw3,Exymax_tw1,Exymax_tw2,Exymax_tw3,Eyymax_tf,TFMmax_tw1,TFMmax_tw2,TFMmax_tw3,TFMmax_frame tw1,tw2,tw3,Exymax_tw1,Exymax_tw2,Exymax_tw3,Eyymax_tf,TFMmax_tw1,TFMmax_tw2,TFMmax_tw3,TFMmax_frame
8.93,10.69,11.92,0.0314285,0.036188,0.0500572,0.096886,60.3168,73.0743,75.173,70.8353 8.93,10.69,11.92,0.0314285,0.0361880,0.0500572,0.0968860,60.3168,73.0743,75.1730,70.8353
14.65,7.11,11.38,0.0041545,0.0770628,0.0446173,0.1298872,5.7242,140.3068,76.6369,85.1589 14.65,7.11,11.38,0.0041545,0.0770628,0.0446173,0.1298872,5.7242,140.3068,76.6369,85.1589
6.29,6.52,9.56,0.042533,0.0678026,0.0470426,0.1058771,91.4702,145.2878,86.7668,68.0555 6.29,6.52,9.56,0.0425330,0.0678026,0.0470426,0.1058771,91.4702,145.2878,86.7668,68.0555
4.58,8.02,12.59,0.1047867,0.0870891,0.0314758,0.083459,202.6256,124.2751,48.3992,57.2506 4.58,8.02,12.59,0.1047867,0.0870891,0.0314758,0.0834590,202.6256,124.2751,48.3992,57.2506
10.95,6.12,10.09,0.0169834,0.0839863,0.0570973,0.1329752,18.5951,160.5635,88.2842,83.5555 10.95,6.12,10.09,0.0169834,0.0839863,0.0570973,0.1329752,18.5951,160.5635,88.2842,83.5555
13.03,8.96,14.9,0.0109994,0.0652393,0.0298515,0.116083,14.5435,117.7404,55.5012,82.8776 13.03,8.96,14.90,0.0109994,0.0652393,0.0298515,0.1160830,14.5435,117.7404,55.5012,82.8776
15.45,11.1,6.67,0.001837,0.0306984,0.1385195,0.2498688,1.1291,55.9504,197.4358,125.0097 15.45,11.10,6.67,0.0018370,0.0306984,0.1385195,0.2498688,1.1291,55.9504,197.4358,125.0097
14.25,13.56,5.6,0.0018503,0.032009,0.1733572,0.3333523,0.8401,48.5556,257.8822,152.0872 14.25,13.56,5.60,0.0018503,0.0320090,0.1733572,0.3333523,0.8401,48.5556,257.8822,152.0872
5.22,14.05,8.76,0.0767186,0.0184562,0.0625321,0.1078929,170.6635,33.4413,105.7267,80.4824 5.22,14.05,8.76,0.0767186,0.0184562,0.0625321,0.1078929,170.6635,33.4413,105.7267,80.4824
12.69,4.24,7.49,0.0087924,0.1325386,0.1181118,0.1384105,9.2216,229.5643,119.7812,81.6469 12.69,4.24,7.49,0.0087924,0.1325386,0.1181118,0.1384105,9.2216,229.5643,119.7812,81.6469
7.82,9.84,14.42,0.042177,0.047482,0.0297936,0.0934205,88.3429,98.3484,53.4841,67.0666 7.82,9.84,14.42,0.0421770,0.0474820,0.0297936,0.0934205,88.3429,98.3484,53.4841,67.0666
5.72,11.78,13.37,0.0772789,0.0353789,0.0293972,0.0719482,166.3376,67.8034,55.0039,54.502 5.72,11.78,13.37,0.0772789,0.0353789,0.0293972,0.0719482,166.3376,67.8034,55.0039,54.5020
10.54,4.85,8.21,0.0161377,0.1107135,0.0903302,0.1383831,18.5385,195.3177,110.0121,81.7428 10.54,4.85,8.21,0.0161377,0.1107135,0.0903302,0.1383831,18.5385,195.3177,110.0121,81.7428
7.07,15.26,4.16,0.0205172,0.0282205,0.1826803,0.28507,40.8943,38.3046,285.025,149.7966 7.07,15.26,4.16,0.0205172,0.0282205,0.1826803,0.2850700,40.8943,38.3046,285.0250,149.7966
11.94,14.8,15.93,0.0239298,0.0817664,0.0338733,0.1028892,42.4723,87.8776,64.2093,69.8335 11.94,14.80,15.93,0.0239298,0.0817664,0.0338733,0.1028892,42.4723,87.8776,64.2093,69.8335
9.81,12.65,4.96,0.0089586,0.032101,0.1749998,0.3156683,13.1979,47.4165,266.644,147.5593 9.81,12.65,4.96,0.0089586,0.0321010,0.1749998,0.3156683,13.1979,47.4165,266.6440,147.5593
7.2,9.27,9.81,0.0394869,0.0424188,0.056382,0.092518,82.3909,85.2522,90.2833,70.8243 7.21,9.27,9.82,0.0389181,0.0460074,0.0552465,0.0930625,82.2030,85.0636,90.3346,70.8627
Code/models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/cv_detailed_per_output_B29_H30.csv
output,model,cv_rmse,std_rmse,cv_rmse_dispersion,cv_mae,cv_r2,BEST_PARAMS,fit_time_sec,gpr_kernel output,model,cv_rmse,std_rmse,cv_rmse_dispersion,cv_mae,cv_r2,BEST_PARAMS,fit_time_sec,gpr_kernel
exymax_tw1,SVR,0.00396466323858515,0.002615136403582738,0.65961123207932,0.00396466323858515,,"{""svr__C"": 6378.353975271747, ""svr__epsilon"": 0.00017612867304089586, ""svr__gamma"": 0.0001299730411455674}",17.99, exymax_tw1,SVR,0.004061647396259379,0.00242300813481289,0.5965579722760738,0.004061647396259379,,"{""svr__C"": 9561.792183843145, ""svr__epsilon"": 0.0003108623718211416, ""svr__gamma"": 0.00010040004983034561}",17.54,
exymax_tw1,FlexibleMLP,0.006146755869837457,0.004321525502374639,0.7030579372089031,0.006146755869837457,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.0013279914857172704, ""mlp__learning_rate_init"": 0.003925505663369577, ""mlp__n_layers"": 5, ""mlp__n_neurons"": 271}",207.91, exymax_tw1,FlexibleMLP,0.005571216562187513,0.003812465998428609,0.6843148091395789,0.005571216562187513,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 3.278043287004692e-05, ""mlp__learning_rate_init"": 0.001570703295827246, ""mlp__n_layers"": 4, ""mlp__n_neurons"": 298}",162.65,
exymax_tw1,GaussianProcess,0.007876513719748384,0.00434651863660955,0.5518328020824421,0.007876513719748384,,"{""gpr__amplitude"": 61.29452020123816, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 9.283042242111568, ""gpr__n_restarts_optimizer"": 2, ""gpr__noise"": 3.341230773727593e-07, ""gpr__rq_alpha"": 20.45913485886371}",22.41,"2.23**2 * Matern(length_scale=4.84, nu=1.5) + WhiteKernel(noise_level=3.34e-07)" exymax_tw1,GradientBoosting,0.0073528745111824095,0.006233050109900674,0.8477025006235749,0.0073528745111824095,,"{""learning_rate"": 0.07999999999999999, ""max_depth"": 1, ""max_features"": 0.971125246014092, ""n_estimators"": 1013, ""subsample"": 0.7139567346576897}",33.42,
exymax_tw1,GradientBoosting,0.008213443438915214,0.007646993125911322,0.9310337598089424,0.008213443438915214,,"{""learning_rate"": 0.07633957742290184, ""max_depth"": 1, ""max_features"": 0.6829045835646892, ""n_estimators"": 272, ""subsample"": 0.8571124953034719}",26.31, exymax_tw1,GaussianProcess,0.007876521279568942,0.004346524198479995,0.5518329785706955,0.007876521279568942,,"{""gpr__amplitude"": 100.0, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 1.6377456535597703, ""gpr__n_restarts_optimizer"": 1, ""gpr__noise"": 9.489819784743017e-12, ""gpr__rq_alpha"": 0.36165095086180327}",34.15,"2.23**2 * Matern(length_scale=4.84, nu=1.5) + WhiteKernel(noise_level=9.49e-12)"
exymax_tw1,XGBoost,0.013517901650694013,0.007837181706069635,0.5797631843006693,0.013517901650694013,,"{""colsample_bytree"": 0.9579303217354647, ""learning_rate"": 0.03023110137240505, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 253, ""subsample"": 0.9457514346290097}",20.26, exymax_tw1,RandomForest,0.012349746678461733,0.004361825360485833,0.3531914843316556,0.012349746678461733,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 217}",34.44,
exymax_tw1,RandomForest,0.01586775883475884,0.00676462747445339,0.4263127228550546,0.01586775883475884,,"{""max_depth"": 2, ""max_features"": 0.5885114671358112, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 309}",36.97, exymax_tw1,XGBoost,0.012719780605179072,0.007866125346820437,0.6184167471896184,0.012719780605179072,,"{""colsample_bytree"": 1.0, ""learning_rate"": 0.01, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 1200, ""subsample"": 0.7}",23.32,
exymax_tw2,SVR,0.003944696026097462,0.0033326552556185705,0.8448446302504097,0.003944696026097462,,"{""svr__C"": 1776.5766649807683, ""svr__epsilon"": 0.00032780432870046914, ""svr__gamma"": 0.006225026900894044}",18.27, exymax_tw2,SVR,0.0038124872313874763,0.003340839390048544,0.8762886764692756,0.0038124872313874763,,"{""svr__C"": 9978.136366417832, ""svr__epsilon"": 0.00033424143012809605, ""svr__gamma"": 0.001381610799038026}",17.31,
exymax_tw2,GaussianProcess,0.004117209548944661,0.003968367538203973,0.9638488133840942,0.004117209548944661,,"{""gpr__amplitude"": 17.76576664980768, ""gpr__kernel_type"": ""RBF"", ""gpr__length_scale"": 2.467108843522573, ""gpr__n_restarts_optimizer"": 4, ""gpr__noise"": 2.3000512910597964e-12, ""gpr__rq_alpha"": 0.024337729656929853}",23.73,2.6**2 * RBF(length_scale=3.89) + WhiteKernel(noise_level=2.3e-12) exymax_tw2,GaussianProcess,0.00411721019435559,0.003968373433341647,0.9638500941200456,0.00411721019435559,,"{""gpr__amplitude"": 0.01745144512080451, ""gpr__kernel_type"": ""RBF"", ""gpr__length_scale"": 0.48474716099273013, ""gpr__n_restarts_optimizer"": 6, ""gpr__noise"": 1.990294785454531e-12, ""gpr__rq_alpha"": 14.396225639189552}",32.4,2.6**2 * RBF(length_scale=3.89) + WhiteKernel(noise_level=1.99e-12)
exymax_tw2,FlexibleMLP,0.00434275413869013,0.003983717060509943,0.9173250276865823,0.00434275413869013,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 0.001393532504374869, ""mlp__learning_rate_init"": 0.00020118052438201193, ""mlp__n_layers"": 1, ""mlp__n_neurons"": 451}",127.32, exymax_tw2,FlexibleMLP,0.0055091952924174155,0.004769524504816916,0.8657388695916179,0.0055091952924174155,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 1e-05, ""mlp__learning_rate_init"": 0.0004153494678254136, ""mlp__n_layers"": 5, ""mlp__n_neurons"": 512}",209.43,
exymax_tw2,GradientBoosting,0.011558255569988348,0.011762883165475899,1.017704020667174,0.011558255569988348,,"{""learning_rate"": 0.060830170099953736, ""max_depth"": 1, ""max_features"": 0.9681857376712572, ""n_estimators"": 486, ""subsample"": 0.8569374216814941}",26.8, exymax_tw2,GradientBoosting,0.011707831856408241,0.011471790090814105,0.9798389856901695,0.011707831856408241,,"{""learning_rate"": 0.04631463751979753, ""max_depth"": 1, ""max_features"": 0.9620112633855968, ""n_estimators"": 1126, ""subsample"": 0.7655177807199793}",31.41,
exymax_tw2,XGBoost,0.014543694154083728,0.01196044827188117,0.8223803488416176,0.014543694154083728,,"{""colsample_bytree"": 0.9794606658475393, ""learning_rate"": 0.011326363591482994, ""max_depth"": 2, ""min_child_weight"": 3, ""n_estimators"": 668, ""subsample"": 0.7154066006870664}",19.74, exymax_tw2,RandomForest,0.014368855743601206,0.011847255148081014,0.8245092970160048,0.014368855743601206,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 200}",38.05,
exymax_tw2,RandomForest,0.02060285611309525,0.012133997872073855,0.588947367562376,0.02060285611309525,,"{""max_depth"": 4, ""max_features"": 0.7312909781953587, ""min_samples_leaf"": 2, ""min_samples_split"": 4, ""n_estimators"": 205}",30.21, exymax_tw2,XGBoost,0.014855466958791017,0.009169871468427567,0.6172725161628874,0.014855466958791017,,"{""colsample_bytree"": 1.0, ""learning_rate"": 0.01, ""max_depth"": 1, ""min_child_weight"": 3, ""n_estimators"": 782, ""subsample"": 0.7}",26.59,
tfmmax_frame,SVR,1.8448767818749428,1.8179078246426847,0.985381702725507,1.8448767818749428,,"{""svr__C"": 872.0933090674746, ""svr__epsilon"": 0.008696229680912925, ""svr__gamma"": 0.014144889231893125}",17.4, tfmmax_frame,SVR,1.8473828812332656,1.8543467263702917,1.0037695732745868,1.8473828812332656,,"{""svr__C"": 2619.206360822925, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.008488619625011348}",19.44,
tfmmax_frame,GaussianProcess,2.6483258733034294,2.28636740402283,0.8633255548611528,2.6483258733034294,,"{""gpr__amplitude"": 15.783879853890564, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 1.2778531518898433, ""gpr__n_restarts_optimizer"": 4, ""gpr__noise"": 3.279409439150647e-05, ""gpr__rq_alpha"": 3.0182479639711155}",22.69,"2.82**2 * Matern(length_scale=6.88, nu=1.5) + WhiteKernel(noise_level=3.28e-05)" tfmmax_frame,GaussianProcess,2.677908229833566,2.259126440914181,0.8436160790523383,2.677908229833566,,"{""gpr__amplitude"": 15.783879853890564, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 1.2778531518898433, ""gpr__n_restarts_optimizer"": 7, ""gpr__noise"": 1.7014397704982245e-05, ""gpr__rq_alpha"": 7.38115974661544}",32.29,"2.82**2 * Matern(length_scale=6.88, nu=1.5) + WhiteKernel(noise_level=1.7e-05)"
tfmmax_frame,GradientBoosting,3.1518436479631493,2.5683108924816773,0.814859865952242,3.1518436479631493,,"{""learning_rate"": 0.07759704098928578, ""max_depth"": 1, ""max_features"": 0.799263994750068, ""n_estimators"": 1478, ""subsample"": 0.801796823725195}",28.91, tfmmax_frame,GradientBoosting,3.1449430403648044,2.5970563426620688,0.8257880379165207,3.1449430403648044,,"{""learning_rate"": 0.07901288854741141, ""max_depth"": 1, ""max_features"": 0.9043798942310827, ""n_estimators"": 1368, ""subsample"": 0.7891742230083864}",36.02,
tfmmax_frame,XGBoost,5.728780969238281,4.60997740521154,0.8047047757569441,5.728780969238281,,"{""colsample_bytree"": 0.8170058789463338, ""learning_rate"": 0.048168492312858205, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 219, ""subsample"": 0.7217020855465729}",19.57, tfmmax_frame,XGBoost,5.7461626373291015,4.399980671230776,0.7657250497309191,5.7461626373291015,,"{""colsample_bytree"": 0.7887391540243318, ""learning_rate"": 0.01, ""max_depth"": 1, ""min_child_weight"": 3, ""n_estimators"": 1175, ""subsample"": 0.7}",24.73,
tfmmax_frame,RandomForest,8.330089468678585,5.641153601590913,0.6772020424032467,8.330089468678585,,"{""max_depth"": 3, ""max_features"": 0.9572412447620999, ""min_samples_leaf"": 2, ""min_samples_split"": 3, ""n_estimators"": 271}",29.44, tfmmax_frame,RandomForest,6.650267066145972,4.655061953713272,0.6999812048767868,6.650267066145972,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 800}",39.2,
tfmmax_frame,FlexibleMLP,9.45273246558955,6.159544863663379,0.6516152748515579,9.45273246558955,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 1.546571475369124e-05, ""mlp__learning_rate_init"": 0.001352838933163239, ""mlp__n_layers"": 2, ""mlp__n_neurons"": 199}",207.83, tfmmax_frame,FlexibleMLP,7.551392118131586,7.5526958607485355,1.000172649307115,7.551392118131586,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 0.01, ""mlp__learning_rate_init"": 0.0001, ""mlp__n_layers"": 3, ""mlp__n_neurons"": 502}",2458.03,
tfmmax_tw1,SVR,5.999344323594288,7.75478808008392,1.2926059352162542,5.999344323594288,,"{""svr__C"": 9473.450684118796, ""svr__epsilon"": 0.03698445205342322, ""svr__gamma"": 0.017296157005375526}",16.83, tfmmax_tw1,SVR,5.76420278175457,7.84041278534167,1.3601903129013655,5.76420278175457,,"{""svr__C"": 10000.0, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.01599115973912198}",17.62,
tfmmax_tw1,FlexibleMLP,10.092645500012402,8.332835282086247,0.8256343970543707,10.092645500012402,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.005484838629875449, ""mlp__learning_rate_init"": 0.008880039744756154, ""mlp__n_layers"": 4, ""mlp__n_neurons"": 215}",106.88, tfmmax_tw1,FlexibleMLP,8.30516696377768,5.34347989931817,0.6433922307189403,8.30516696377768,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.004356255540694861, ""mlp__learning_rate_init"": 0.005440491666387846, ""mlp__n_layers"": 2, ""mlp__n_neurons"": 364}",142.88,
tfmmax_tw1,GradientBoosting,13.959636826053607,14.300194228658684,1.0243958640793203,13.959636826053607,,"{""learning_rate"": 0.07549905079959124, ""max_depth"": 1, ""max_features"": 0.6987972259552133, ""n_estimators"": 1468, ""subsample"": 0.7071168121432537}",28.75, tfmmax_tw1,GradientBoosting,14.037351115199302,14.327828368857423,1.0206931671990167,14.037351115199302,,"{""learning_rate"": 0.0680543874152698, ""max_depth"": 1, ""max_features"": 0.7103120195166182, ""n_estimators"": 1500, ""subsample"": 0.7}",41.08,
tfmmax_tw1,GaussianProcess,14.306704291634716,11.044424957912446,0.7719754831565394,14.306704291634716,,"{""gpr__amplitude"": 18.46745783020663, ""gpr__kernel_type"": ""Matern52"", ""gpr__length_scale"": 5.716255151488844, ""gpr__n_restarts_optimizer"": 4, ""gpr__noise"": 5.890882642108198e-06, ""gpr__rq_alpha"": 20.258671252175002}",23.62,"2**2 * Matern(length_scale=3.31, nu=2.5) + WhiteKernel(noise_level=7.4e-10)" tfmmax_tw1,GaussianProcess,14.579264561521235,11.004017114326576,0.7547717560026492,14.579264561521235,,"{""gpr__amplitude"": 100.0, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 1.3014302820144033, ""gpr__n_restarts_optimizer"": 7, ""gpr__noise"": 6.791609124669547e-05, ""gpr__rq_alpha"": 1.1811151895437066}",35.87,"2.03**2 * Matern(length_scale=4.35, nu=1.5) + WhiteKernel(noise_level=3.7e-09)"
tfmmax_tw1,RandomForest,34.74053457854516,14.312277125903012,0.4119763066266084,34.74053457854516,,"{""max_depth"": 3, ""max_features"": 0.9287094401669573, ""min_samples_leaf"": 2, ""min_samples_split"": 4, ""n_estimators"": 317}",28.79, tfmmax_tw1,XGBoost,26.66287326660156,17.10121411033928,0.6413867680105054,26.66287326660156,,"{""colsample_bytree"": 1.0, ""learning_rate"": 0.04920876654372351, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 200, ""subsample"": 0.7}",27.04,
tfmmax_tw1,XGBoost,38.75318451690674,21.75531074251873,0.5613812390831417,38.75318451690674,,"{""colsample_bytree"": 0.8230311876559941, ""learning_rate"": 0.08846938749167613, ""max_depth"": 4, ""min_child_weight"": 8, ""n_estimators"": 434, ""subsample"": 0.7101294823253874}",20.17, tfmmax_tw1,RandomForest,27.63103577794101,10.109773859992078,0.36588472257211313,27.63103577794101,,"{""max_depth"": 3, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 3, ""n_estimators"": 389}",43.93,
tfmmax_tw2,FlexibleMLP,10.397110371014389,7.944933680445426,0.7641482485936409,10.397110371014389,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.0013843697688682688, ""mlp__learning_rate_init"": 0.0012259398541783753, ""mlp__n_layers"": 1, ""mlp__n_neurons"": 68}",109.38, tfmmax_tw2,SVR,8.5556547519197,5.490513935936841,0.641740941534006,8.5556547519197,,"{""svr__C"": 7123.257292646505, ""svr__epsilon"": 0.1, ""svr__gamma"": 0.012177597403964988}",20.86,
tfmmax_tw2,SVR,11.550861893053522,12.144401372111433,1.0513848650043036,11.550861893053522,,"{""svr__C"": 1390.8696978218766, ""svr__epsilon"": 0.0010859185217669352, ""svr__gamma"": 0.003365450134654306}",17.35, tfmmax_tw2,FlexibleMLP,10.410474638254907,7.194404117616944,0.6910735934344423,10.410474638254907,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.01, ""mlp__learning_rate_init"": 0.009999866980900942, ""mlp__n_layers"": 2, ""mlp__n_neurons"": 410}",225.34,
tfmmax_tw2,GaussianProcess,12.249019606125456,11.640315522790269,0.9503058936218219,12.249019606125456,,"{""gpr__amplitude"": 0.27391266066819153, ""gpr__kernel_type"": ""Matern52"", ""gpr__length_scale"": 1.4509550508588704, ""gpr__n_restarts_optimizer"": 10, ""gpr__noise"": 5.139540511975707e-05, ""gpr__rq_alpha"": 0.024884673238486348}",21.69,"2.88**2 * Matern(length_scale=5.43, nu=2.5) + WhiteKernel(noise_level=0.01)" tfmmax_tw2,GradientBoosting,12.412813395497732,9.644252530649691,0.7769594388769084,12.412813395497732,,"{""learning_rate"": 0.0152095788809544, ""max_depth"": 1, ""max_features"": 1.0, ""n_estimators"": 1500, ""subsample"": 0.7799913032638333}",37.26,
tfmmax_tw2,GradientBoosting,15.007948175448531,13.75052357973865,0.9162160889010198,15.007948175448531,,"{""learning_rate"": 0.033771951671657495, ""max_depth"": 1, ""max_features"": 0.9473733636619561, ""n_estimators"": 1228, ""subsample"": 0.8237382439016434}",26.53, tfmmax_tw2,GaussianProcess,12.8960010364433,12.020235809862935,0.9320901708905337,12.8960010364433,,"{""gpr__amplitude"": 64.11805807786253, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 0.8700053838385312, ""gpr__n_restarts_optimizer"": 10, ""gpr__noise"": 0.0001, ""gpr__rq_alpha"": 89.63083374201149}",35.91,"2.88**2 * Matern(length_scale=7.29, nu=1.5) + WhiteKernel(noise_level=0.01)"
tfmmax_tw2,XGBoost,25.2020942855835,17.44008847232459,0.6920094923341726,25.2020942855835,,"{""colsample_bytree"": 0.7483276113437526, ""learning_rate"": 0.03961468376647114, ""max_depth"": 1, ""min_child_weight"": 3, ""n_estimators"": 220, ""subsample"": 0.7041554603114548}",18.93, tfmmax_tw2,RandomForest,20.37966172767856,17.52666810815926,0.8600078029928967,20.37966172767856,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 3, ""n_estimators"": 200}",40.3,
tfmmax_tw2,RandomForest,29.43283981917171,20.485769517263364,0.6960174296168159,29.43283981917171,,"{""max_depth"": 4, ""max_features"": 0.9902246158537406, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 211}",28.41, tfmmax_tw2,XGBoost,24.740090702819828,14.52448238931494,0.5870828269703744,24.740090702819828,,"{""colsample_bytree"": 1.0, ""learning_rate"": 0.01, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 1200, ""subsample"": 0.7}",24.8,
Code/models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/cv_summary_per_output_B29_H30.csv
output,best_model,cv_rmse,std_rmse,cv_rmse_dispersion,cv_mae,cv_r2,BEST_PARAMS,model_path,train_time_sec,gpr_kernel,selected_by output,best_model,cv_rmse,std_rmse,cv_rmse_dispersion,cv_mae,cv_r2,BEST_PARAMS,model_path,train_time_sec,gpr_kernel,selected_by
exymax_tw2,SVR,0.003944696026097462,0.0033326552556185705,0.8448446302504097,0.003944696026097462,,"{""svr__C"": 1776.5766649807683, ""svr__epsilon"": 0.00032780432870046914, ""svr__gamma"": 0.006225026900894044}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_exymax_tw2.joblib,246.08,,lowest_cv_rmse_dispersion_within_5pct_rmse_band exymax_tw2,SVR,0.0038124872313874763,0.003340839390048544,0.8762886764692756,0.0038124872313874763,,"{""svr__C"": 9978.136366417832, ""svr__epsilon"": 0.00033424143012809605, ""svr__gamma"": 0.001381610799038026}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_exymax_tw2.joblib,355.19,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
exymax_tw1,SVR,0.00396466323858515,0.002615136403582738,0.65961123207932,0.00396466323858515,,"{""svr__C"": 6378.353975271747, ""svr__epsilon"": 0.00017612867304089586, ""svr__gamma"": 0.0001299730411455674}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_exymax_tw1.joblib,331.85,,lowest_cv_rmse_dispersion_within_5pct_rmse_band exymax_tw1,SVR,0.004061647396259379,0.00242300813481289,0.5965579722760738,0.004061647396259379,,"{""svr__C"": 9561.792183843145, ""svr__epsilon"": 0.0003108623718211416, ""svr__gamma"": 0.00010040004983034561}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_exymax_tw1.joblib,305.52,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
tfmmax_frame,SVR,1.8448767818749428,1.8179078246426847,0.985381702725507,1.8448767818749428,,"{""svr__C"": 872.0933090674746, ""svr__epsilon"": 0.008696229680912925, ""svr__gamma"": 0.014144889231893125}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_tfmmax_frame.joblib,325.87,,lowest_cv_rmse_dispersion_within_5pct_rmse_band tfmmax_frame,SVR,1.8473828812332656,1.8543467263702917,1.0037695732745868,1.8473828812332656,,"{""svr__C"": 2619.206360822925, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.008488619625011348}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_tfmmax_frame.joblib,2609.71,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
tfmmax_tw1,SVR,5.999344323594288,7.75478808008392,1.2926059352162542,5.999344323594288,,"{""svr__C"": 9473.450684118796, ""svr__epsilon"": 0.03698445205342322, ""svr__gamma"": 0.017296157005375526}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_tfmmax_tw1.joblib,225.04,,lowest_cv_rmse_dispersion_within_5pct_rmse_band tfmmax_tw1,SVR,5.76420278175457,7.84041278534167,1.3601903129013655,5.76420278175457,,"{""svr__C"": 10000.0, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.01599115973912198}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_tfmmax_tw1.joblib,308.43,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
tfmmax_tw2,FlexibleMLP,10.397110371014389,7.944933680445426,0.7641482485936409,10.397110371014389,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.0013843697688682688, ""mlp__learning_rate_init"": 0.0012259398541783753, ""mlp__n_layers"": 1, ""mlp__n_neurons"": 68}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_tfmmax_tw2.joblib,222.32,,lowest_cv_rmse_dispersion_within_5pct_rmse_band tfmmax_tw2,SVR,8.5556547519197,5.490513935936841,0.641740941534006,8.5556547519197,,"{""svr__C"": 7123.257292646505, ""svr__epsilon"": 0.1, ""svr__gamma"": 0.012177597403964988}",../../models/width_optimization/2W/ml_models/per_output_models_B29_H30/it0/best_model_tfmmax_tw2.joblib,384.47,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
Code/models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/cv_detailed_per_output_B34_H30.csv
output,model,cv_rmse,std_rmse,cv_rmse_dispersion,cv_mae,cv_r2,BEST_PARAMS,fit_time_sec,gpr_kernel output,model,cv_rmse,std_rmse,cv_rmse_dispersion,cv_mae,cv_r2,BEST_PARAMS,fit_time_sec,gpr_kernel
exymax_tw1,SVR,0.0008167425561569747,0.0006767801130116837,0.8286333409589217,0.0008167425561569747,,"{""svr__C"": 4442.8726934917495, ""svr__epsilon"": 0.00012917724146922354, ""svr__gamma"": 0.001504160883442332}",19.98, exymax_tw1,SVR,0.0009391988885412684,0.0008408212014288188,0.8952536163397228,0.0009391988885412684,,"{""svr__C"": 1145.053088286465, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.00290506172148416}",13.58,
exymax_tw1,GaussianProcess,0.006286392345538525,0.0071155636220117235,1.1318993837636724,0.006286392345538525,,"{""gpr__amplitude"": 17.76576664980768, ""gpr__kernel_type"": ""RBF"", ""gpr__length_scale"": 2.467108843522573, ""gpr__n_restarts_optimizer"": 4, ""gpr__noise"": 2.3000512910597964e-12, ""gpr__rq_alpha"": 0.024337729656929853}",23.18,3.73**2 * RBF(length_scale=4.06) + WhiteKernel(noise_level=2.3e-12) exymax_tw1,GaussianProcess,0.006286395530318298,0.007115554222289232,1.1318973150785743,0.006286395530318298,,"{""gpr__amplitude"": 100.0, ""gpr__kernel_type"": ""RBF"", ""gpr__length_scale"": 2.361364647456093, ""gpr__n_restarts_optimizer"": 7, ""gpr__noise"": 8.461979957381341e-11, ""gpr__rq_alpha"": 0.09573383400652338}",32.85,3.73**2 * RBF(length_scale=4.06) + WhiteKernel(noise_level=8.46e-11)
exymax_tw1,FlexibleMLP,0.006347635568669565,0.006781582920172903,1.0683636208803793,0.006347635568669565,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 0.0057547661357822905, ""mlp__learning_rate_init"": 0.0009838602333915323, ""mlp__n_layers"": 5, ""mlp__n_neurons"": 131}",150.05, exymax_tw1,GradientBoosting,0.007213015243224071,0.005835291818934489,0.8089947992853863,0.007213015243224071,,"{""learning_rate"": 0.07999999999999999, ""max_depth"": 1, ""max_features"": 1.0, ""n_estimators"": 888, ""subsample"": 0.7}",27.93,
exymax_tw1,GradientBoosting,0.008566895564146934,0.006986363143533339,0.8155069816389223,0.008566895564146934,,"{""learning_rate"": 0.07755795475040612, ""max_depth"": 1, ""max_features"": 0.8979806010110151, ""n_estimators"": 244, ""subsample"": 0.7027055333399309}",26.43, exymax_tw1,FlexibleMLP,0.007529932161866227,0.0062093826735622245,0.8246266420577266,0.007529932161866227,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 0.0002061045404501547, ""mlp__learning_rate_init"": 0.0011304216699488043, ""mlp__n_layers"": 4, ""mlp__n_neurons"": 469}",160.17,
exymax_tw1,XGBoost,0.021034819270297885,0.014992332658813726,0.712738838692262,0.021034819270297885,,"{""colsample_bytree"": 0.7372214089779456, ""learning_rate"": 0.18729267358417406, ""max_depth"": 3, ""min_child_weight"": 3, ""n_estimators"": 321, ""subsample"": 0.7126363562411662}",21.89, exymax_tw1,RandomForest,0.017197718448375754,0.010107123487360576,0.5877014161907709,0.017197718448375754,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 483}",35.91,
exymax_tw1,RandomForest,0.023420028504697467,0.013801082021962839,0.5892854493834065,0.023420028504697467,,"{""max_depth"": 4, ""max_features"": 0.9153682682439178, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 351}",35.69, exymax_tw1,XGBoost,0.02110078880637884,0.014003884175399965,0.6636663825177257,0.02110078880637884,,"{""colsample_bytree"": 1.0, ""learning_rate"": 0.01, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 200, ""subsample"": 0.7}",19.44,
exymax_tw2,SVR,0.006136161710932672,0.005656049157920998,0.9217568611080983,0.006136161710932672,,"{""svr__C"": 1776.5766649807683, ""svr__epsilon"": 0.00032780432870046914, ""svr__gamma"": 0.006225026900894044}",19.53, exymax_tw2,SVR,0.006029466121502341,0.005518640222413557,0.9152784195491087,0.006029466121502341,,"{""svr__C"": 10000.0, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.0034683155294780777}",15.46,
exymax_tw2,GaussianProcess,0.006934736426116973,0.006580766406441501,0.9489569613140044,0.006934736426116973,,"{""gpr__amplitude"": 0.4369339947510315, ""gpr__kernel_type"": ""Matern52"", ""gpr__length_scale"": 53.88550972627239, ""gpr__n_restarts_optimizer"": 10, ""gpr__noise"": 3.3825346762279944e-05, ""gpr__rq_alpha"": 0.014523514317963207}",24.76,"4.43**2 * Matern(length_scale=7.85, nu=2.5) + WhiteKernel(noise_level=2.91e-12)" exymax_tw2,GaussianProcess,0.0069347318929677704,0.006580766519542526,0.9489575979448915,0.0069347318929677704,,"{""gpr__amplitude"": 1.0862274576850792, ""gpr__kernel_type"": ""Matern52"", ""gpr__length_scale"": 4.174350684925192, ""gpr__n_restarts_optimizer"": 9, ""gpr__noise"": 1.1255372473930115e-11, ""gpr__rq_alpha"": 0.013817053968571608}",28.79,"4.43**2 * Matern(length_scale=7.85, nu=2.5) + WhiteKernel(noise_level=1.13e-11)"
exymax_tw2,FlexibleMLP,0.010845815935089018,0.005846310882948516,0.539038364465894,0.010845815935089018,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.001955165749106793, ""mlp__learning_rate_init"": 0.008962267999175521, ""mlp__n_layers"": 2, ""mlp__n_neurons"": 318}",131.61, exymax_tw2,FlexibleMLP,0.009269438568073593,0.007419412657431604,0.800416616707082,0.009269438568073593,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.0015246748254295628, ""mlp__learning_rate_init"": 0.007340675018434775, ""mlp__n_layers"": 2, ""mlp__n_neurons"": 364}",135.05,
exymax_tw2,GradientBoosting,0.016576780300091604,0.016779289923991116,1.012216463042488,0.016576780300091604,,"{""learning_rate"": 0.07819794579224973, ""max_depth"": 1, ""max_features"": 0.8765197496439208, ""n_estimators"": 438, ""subsample"": 0.8561842190690474}",26.43, exymax_tw2,GradientBoosting,0.016700477007043617,0.016547519016267307,0.9908411004840282,0.016700477007043617,,"{""learning_rate"": 0.07999999999999999, ""max_depth"": 1, ""max_features"": 0.7031297650225502, ""n_estimators"": 756, ""subsample"": 0.7620444853070656}",28.18,
exymax_tw2,XGBoost,0.022396076684969665,0.017016238284799323,0.7597865699495112,0.022396076684969665,,"{""colsample_bytree"": 0.7661722145199127, ""learning_rate"": 0.011237187033583663, ""max_depth"": 1, ""min_child_weight"": 3, ""n_estimators"": 574, ""subsample"": 0.7054919534044021}",19.17, exymax_tw2,RandomForest,0.022286896378906268,0.015043413620931024,0.6749891669604147,0.022286896378906268,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 800}",33.65,
exymax_tw2,RandomForest,0.027304968441061936,0.01917816481011502,0.702369052412981,0.027304968441061936,,"{""max_depth"": 2, ""max_features"": 0.752170940676494, ""min_samples_leaf"": 2, ""min_samples_split"": 4, ""n_estimators"": 752}",34.42, exymax_tw2,XGBoost,0.023042750971347094,0.012678377658596918,0.5502111130030466,0.023042750971347094,,"{""colsample_bytree"": 0.7, ""learning_rate"": 0.01, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 1200, ""subsample"": 0.7}",21.28,
tfmmax_frame,SVR,3.715373578591194,2.334838394610003,0.6284262794093868,3.715373578591194,,"{""svr__C"": 5216.1520937153, ""svr__epsilon"": 0.013648761321634565, ""svr__gamma"": 0.015037530613724375}",17.32, tfmmax_frame,SVR,3.2580037673327116,2.5119569360841667,0.7710110593704671,3.2580037673327116,,"{""svr__C"": 10000.0, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.013154596167883838}",20.55,
tfmmax_frame,FlexibleMLP,4.482099086078421,4.649005409680853,1.0372384278876943,4.482099086078421,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 5.449121819350365e-05, ""mlp__learning_rate_init"": 0.0011162805098567702, ""mlp__n_layers"": 1, ""mlp__n_neurons"": 492}",259.62, tfmmax_frame,GaussianProcess,4.861130062071435,2.777104647873538,0.571287871834919,4.861130062071435,,"{""gpr__amplitude"": 24.218605616737108, ""gpr__kernel_type"": ""RQ"", ""gpr__length_scale"": 0.18099022254583827, ""gpr__n_restarts_optimizer"": 10, ""gpr__noise"": 0.0001, ""gpr__rq_alpha"": 0.5177460521210273}",34.18,"2.13**2 * RationalQuadratic(alpha=1e+03, length_scale=3.12) + WhiteKernel(noise_level=0.01)"
tfmmax_frame,GradientBoosting,4.813883516236368,3.768493802179492,0.7828385937194028,4.813883516236368,,"{""learning_rate"": 0.07640636409973982, ""max_depth"": 1, ""max_features"": 0.7555596057831717, ""n_estimators"": 554, ""subsample"": 0.8075333318020734}",24.86, tfmmax_frame,GradientBoosting,4.902889831675344,3.874234194734623,0.7901940136824931,4.902889831675344,,"{""learning_rate"": 0.06640687000693812, ""max_depth"": 1, ""max_features"": 0.6, ""n_estimators"": 1500, ""subsample"": 0.7464304021544185}",35.84,
tfmmax_frame,GaussianProcess,4.83173368974961,2.7611688207228937,0.5714654403616319,4.83173368974961,,"{""gpr__amplitude"": 40.51324264659063, ""gpr__kernel_type"": ""RQ"", ""gpr__length_scale"": 2.479740468116072, ""gpr__n_restarts_optimizer"": 6, ""gpr__noise"": 6.216743947159684e-05, ""gpr__rq_alpha"": 0.05158586777420598}",21.85,"2.13**2 * RationalQuadratic(alpha=1e+03, length_scale=3.12) + WhiteKernel(noise_level=0.01)" tfmmax_frame,XGBoost,6.888140588378905,4.842058044853051,0.7029557516613651,6.888140588378905,,"{""colsample_bytree"": 1.0, ""learning_rate"": 0.0748481134538213, ""max_depth"": 3, ""min_child_weight"": 3, ""n_estimators"": 200, ""subsample"": 0.7}",35.84,
tfmmax_frame,XGBoost,7.445577368164061,3.926935852928922,0.5274185813607669,7.445577368164061,,"{""colsample_bytree"": 0.8886494972499394, ""learning_rate"": 0.06885690196837727, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 223, ""subsample"": 0.7995953415735013}",17.82, tfmmax_frame,RandomForest,8.933791985584499,5.538906901260313,0.6199950603503924,8.933791985584499,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 422}",37.07,
tfmmax_frame,RandomForest,11.447609760794228,6.9278513889043944,0.6051788568676493,11.447609760794228,,"{""max_depth"": 3, ""max_features"": 0.9400414852314116, ""min_samples_leaf"": 2, ""min_samples_split"": 3, ""n_estimators"": 268}",29.09, tfmmax_frame,FlexibleMLP,13.355614585858065,9.490152539058366,0.7105740045169661,13.355614585858065,,"{""mlp__activation"": ""tanh"", ""mlp__alpha"": 0.0025016366510076833, ""mlp__learning_rate_init"": 0.0005251637934360895, ""mlp__n_layers"": 4, ""mlp__n_neurons"": 315}",493.94,
tfmmax_tw1,SVR,6.741286194591464,8.748451173110228,1.2977421400873193,6.741286194591464,,"{""svr__C"": 2768.7251482836446, ""svr__epsilon"": 0.023317529207347875, ""svr__gamma"": 0.028189099645040516}",16.7, tfmmax_tw1,SVR,7.183119170413128,8.309840082499269,1.156856775636835,7.183119170413128,,"{""svr__C"": 8486.611629527324, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.01016134780490442}",24.3,
tfmmax_tw1,FlexibleMLP,8.154980516638027,9.240677786904534,1.1331330305512608,8.154980516638027,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.004391920649227789, ""mlp__learning_rate_init"": 0.0019268504610142117, ""mlp__n_layers"": 1, ""mlp__n_neurons"": 93}",114.84, tfmmax_tw1,FlexibleMLP,11.286075138614564,13.407473120948438,1.1879659630366672,11.286075138614564,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.0018350308720501253, ""mlp__learning_rate_init"": 0.0005762697260565502, ""mlp__n_layers"": 2, ""mlp__n_neurons"": 271}",187.23,
tfmmax_tw1,GaussianProcess,13.738837154661752,12.333199140124707,0.8976887200340609,13.738837154661752,,"{""gpr__amplitude"": 45.10334219573621, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 0.8138746751355537, ""gpr__n_restarts_optimizer"": 7, ""gpr__noise"": 3.723777199450807e-05, ""gpr__rq_alpha"": 7.75507207048533}",23.12,"2.32**2 * Matern(length_scale=5.44, nu=1.5) + WhiteKernel(noise_level=1.24e-09)" tfmmax_tw1,GaussianProcess,13.737825143457947,12.334799542075679,0.8978713452288771,13.737825143457947,,"{""gpr__amplitude"": 16.140832807988044, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 0.7808758275555846, ""gpr__n_restarts_optimizer"": 7, ""gpr__noise"": 6.169251139368679e-05, ""gpr__rq_alpha"": 26.452608826658203}",38.33,"2.32**2 * Matern(length_scale=5.44, nu=1.5) + WhiteKernel(noise_level=8.68e-10)"
tfmmax_tw1,GradientBoosting,17.145239175442043,19.73776492606,1.1512096579166626,17.145239175442043,,"{""learning_rate"": 0.07493152135903128, ""max_depth"": 1, ""max_features"": 0.6532367393810323, ""n_estimators"": 755, ""subsample"": 0.7086631175457405}",29.24, tfmmax_tw1,GradientBoosting,17.125644279942367,19.711357868064262,1.1509848941070377,17.125644279942367,,"{""learning_rate"": 0.07405969628647097, ""max_depth"": 1, ""max_features"": 0.627031854430126, ""n_estimators"": 1500, ""subsample"": 0.7044256922159705}",36.7,
tfmmax_tw1,XGBoost,32.773460178375245,21.207408561487284,0.6470909219247001,32.773460178375245,,"{""colsample_bytree"": 0.8632491169261263, ""learning_rate"": 0.11522234452111198, ""max_depth"": 2, ""min_child_weight"": 3, ""n_estimators"": 207, ""subsample"": 0.7108457019088024}",20.08, tfmmax_tw1,XGBoost,31.923099018096924,20.0219722136703,0.6271938762060669,31.923099018096924,,"{""colsample_bytree"": 1.0, ""learning_rate"": 0.01, ""max_depth"": 4, ""min_child_weight"": 3, ""n_estimators"": 1200, ""subsample"": 0.7}",23.41,
tfmmax_tw1,RandomForest,42.14690362269041,19.094085304139742,0.4530364905349811,42.14690362269041,,"{""max_depth"": 3, ""max_features"": 0.9287094401669573, ""min_samples_leaf"": 2, ""min_samples_split"": 4, ""n_estimators"": 317}",28.74, tfmmax_tw1,RandomForest,34.50481845877975,13.424290144940972,0.38905552165063373,34.50481845877975,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 800}",39.26,
tfmmax_tw2,SVR,11.697291038282572,9.840867849708195,0.8412946055203101,11.697291038282572,,"{""svr__C"": 1174.7546400335007, ""svr__epsilon"": 0.00019247589266324104, ""svr__gamma"": 0.0037290003646388062}",16.76, tfmmax_tw2,SVR,10.383152021375174,7.182550943352127,0.6917505328406864,10.383152021375174,,"{""svr__C"": 10000.0, ""svr__epsilon"": 0.1, ""svr__gamma"": 0.014504091099759808}",17.64,
tfmmax_tw2,GaussianProcess,11.739301648902373,8.161087859526862,0.6951936412921058,11.739301648902373,,"{""gpr__amplitude"": 1.7905875875772057, ""gpr__kernel_type"": ""Matern52"", ""gpr__length_scale"": 1.0810730734949554, ""gpr__n_restarts_optimizer"": 9, ""gpr__noise"": 9.47686335625517e-05, ""gpr__rq_alpha"": 1.4204189514600063}",24.26,"3.2**2 * Matern(length_scale=6, nu=2.5) + WhiteKernel(noise_level=0.00704)" tfmmax_tw2,GaussianProcess,12.568712109000534,10.028846117767344,0.7979215396767362,12.568712109000534,,"{""gpr__amplitude"": 1.3398234241115743, ""gpr__kernel_type"": ""Matern32"", ""gpr__length_scale"": 1.3199024542148905, ""gpr__n_restarts_optimizer"": 9, ""gpr__noise"": 1e-12, ""gpr__rq_alpha"": 0.01}",33.86,"3.25**2 * Matern(length_scale=8.09, nu=1.5) + WhiteKernel(noise_level=1e-12)"
tfmmax_tw2,FlexibleMLP,14.472444964284271,9.499717651642397,0.6564003300814902,14.472444964284271,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.0011872298329982118, ""mlp__learning_rate_init"": 0.0015870562518204246, ""mlp__n_layers"": 1, ""mlp__n_neurons"": 403}",159.84, tfmmax_tw2,FlexibleMLP,15.01279775850158,12.692329287957353,0.8454339752075744,15.01279775850158,,"{""mlp__activation"": ""relu"", ""mlp__alpha"": 0.01, ""mlp__learning_rate_init"": 0.01, ""mlp__n_layers"": 2, ""mlp__n_neurons"": 193}",633.68,
tfmmax_tw2,GradientBoosting,21.105693204101858,16.53654019238579,0.7835108770164412,21.105693204101858,,"{""learning_rate"": 0.05361767415315779, ""max_depth"": 1, ""max_features"": 0.817683854053919, ""n_estimators"": 1282, ""subsample"": 0.855049432657501}",26.91, tfmmax_tw2,GradientBoosting,16.124106409066343,14.362760324592521,0.8907631815501141,16.124106409066343,,"{""learning_rate"": 0.06268027396088106, ""max_depth"": 1, ""max_features"": 1.0, ""n_estimators"": 1500, ""subsample"": 0.7}",34.28,
tfmmax_tw2,XGBoost,28.828188270568848,17.457468069365525,0.6055693790229658,28.828188270568848,,"{""colsample_bytree"": 0.727596261492343, ""learning_rate"": 0.01197861305410859, ""max_depth"": 2, ""min_child_weight"": 3, ""n_estimators"": 1030, ""subsample"": 0.7019121754040568}",20.4, tfmmax_tw2,RandomForest,25.02939522697173,21.87132552293479,0.8738255688801543,25.02939522697173,,"{""max_depth"": 4, ""max_features"": 1.0, ""min_samples_leaf"": 2, ""min_samples_split"": 4, ""n_estimators"": 800}",43.05,
tfmmax_tw2,RandomForest,37.12652917137024,23.078933706718093,0.6216291751967805,37.12652917137024,,"{""max_depth"": 3, ""max_features"": 0.9966276921858434, ""min_samples_leaf"": 2, ""min_samples_split"": 2, ""n_estimators"": 734}",31.4, tfmmax_tw2,XGBoost,28.443999732971193,19.98240260502107,0.7025173250110193,28.443999732971193,,"{""colsample_bytree"": 0.8764302714511444, ""learning_rate"": 0.010883230981476177, ""max_depth"": 1, ""min_child_weight"": 3, ""n_estimators"": 821, ""subsample"": 0.7142781364856831}",23.27,
Code/models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/cv_summary_per_output_B34_H30.csv
output,best_model,cv_rmse,std_rmse,cv_rmse_dispersion,cv_mae,cv_r2,BEST_PARAMS,model_path,train_time_sec,gpr_kernel,selected_by output,best_model,cv_rmse,std_rmse,cv_rmse_dispersion,cv_mae,cv_r2,BEST_PARAMS,model_path,train_time_sec,gpr_kernel,selected_by
exymax_tw1,SVR,0.0008167425561569747,0.0006767801130116837,0.8286333409589217,0.0008167425561569747,,"{""svr__C"": 4442.8726934917495, ""svr__epsilon"": 0.00012917724146922354, ""svr__gamma"": 0.001504160883442332}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_exymax_tw1.joblib,277.24,,lowest_cv_rmse_dispersion_within_5pct_rmse_band exymax_tw1,SVR,0.0009391988885412684,0.0008408212014288188,0.8952536163397228,0.0009391988885412684,,"{""svr__C"": 1145.053088286465, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.00290506172148416}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_exymax_tw1.joblib,289.88,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
exymax_tw2,SVR,0.006136161710932672,0.005656049157920998,0.9217568611080983,0.006136161710932672,,"{""svr__C"": 1776.5766649807683, ""svr__epsilon"": 0.00032780432870046914, ""svr__gamma"": 0.006225026900894044}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_exymax_tw2.joblib,255.91,,lowest_cv_rmse_dispersion_within_5pct_rmse_band exymax_tw2,SVR,0.006029466121502341,0.005518640222413557,0.9152784195491087,0.006029466121502341,,"{""svr__C"": 10000.0, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.0034683155294780777}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_exymax_tw2.joblib,262.41,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
tfmmax_frame,SVR,3.715373578591194,2.334838394610003,0.6284262794093868,3.715373578591194,,"{""svr__C"": 5216.1520937153, ""svr__epsilon"": 0.013648761321634565, ""svr__gamma"": 0.015037530613724375}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_tfmmax_frame.joblib,370.56,,lowest_cv_rmse_dispersion_within_5pct_rmse_band tfmmax_frame,SVR,3.2580037673327116,2.5119569360841667,0.7710110593704671,3.2580037673327116,,"{""svr__C"": 10000.0, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.013154596167883838}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_tfmmax_frame.joblib,657.42,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
tfmmax_tw1,SVR,6.741286194591464,8.748451173110228,1.2977421400873193,6.741286194591464,,"{""svr__C"": 2768.7251482836446, ""svr__epsilon"": 0.023317529207347875, ""svr__gamma"": 0.028189099645040516}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_tfmmax_tw1.joblib,232.73,,lowest_cv_rmse_dispersion_within_5pct_rmse_band tfmmax_tw1,SVR,7.183119170413128,8.309840082499269,1.156856775636835,7.183119170413128,,"{""svr__C"": 8486.611629527324, ""svr__epsilon"": 0.0001, ""svr__gamma"": 0.01016134780490442}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_tfmmax_tw1.joblib,349.24,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
tfmmax_tw2,GaussianProcess,11.739301648902373,8.161087859526862,0.6951936412921058,11.739301648902373,,"{""gpr__amplitude"": 1.7905875875772057, ""gpr__kernel_type"": ""Matern52"", ""gpr__length_scale"": 1.0810730734949554, ""gpr__n_restarts_optimizer"": 9, ""gpr__noise"": 9.47686335625517e-05, ""gpr__rq_alpha"": 1.4204189514600063}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_tfmmax_tw2.joblib,279.57,"3.2**2 * Matern(length_scale=6, nu=2.5) + WhiteKernel(noise_level=0.00704)",lowest_cv_rmse_dispersion_within_5pct_rmse_band tfmmax_tw2,SVR,10.383152021375174,7.182550943352127,0.6917505328406864,10.383152021375174,,"{""svr__C"": 10000.0, ""svr__epsilon"": 0.1, ""svr__gamma"": 0.014504091099759808}",../../models/width_optimization/2W/ml_models/per_output_models_B34_H30/it0/best_model_tfmmax_tw2.joblib,785.79,,lowest_cv_rmse_dispersion_within_5pct_rmse_band
Code/models/width_optimization/2W/rbf_models/per_output_models_B29_H30/it1/rbf_LOO_results.csv 0 → 100644
output,LOO_RMSE,LOO_MAE,LOO_R2,LOO_RMSE_STD,LOO_RMSE_CV,LOO_MAE_STD,LOO_SQERR_STD
exymax_tw1,0.008150172980533048,0.005360493131542719,0.7712544324577968,0.006139253456186208,1.1452777394790479,0.006139253456186208,0.00013861598777410185
exymax_tw2,0.008572862763522579,0.005833395187413924,0.8801428734246546,0.0062821554063591965,1.0769295075213683,0.0062821554063591965,0.00013899408657731815
tfmmax_tw1,10.680222740245128,6.855102191400704,0.9160387068488547,8.18991646640565,1.194718362722497,8.18991646640565,247.11727801028164
tfmmax_tw2,13.395392683237962,11.974619364439299,0.8656018338209511,6.003751844883375,0.501373084368139,6.003751844883375,153.4530221508567
tfmmax_frame,12.261032928851478,7.414418742924263,-0.19112332444358215,9.765209838347236,1.3170566941159592,9.765209838347236,305.4757083982322
Code/models/width_optimization/2W/rbf_models/per_output_models_B34_H30/it1/rbf_LOO_results.csv 0 → 100644
output,LOO_RMSE,LOO_MAE,LOO_R2,LOO_RMSE_STD,LOO_RMSE_CV,LOO_MAE_STD,LOO_SQERR_STD
exymax_tw1,0.0067951309608360955,0.005468506441657539,0.9363332594154856,0.004033514853383247,0.7375898513453452,0.004033514853383247,4.5648407965412274e-05
exymax_tw2,0.010331124979124655,0.007661182112240156,0.9163717254906855,0.006930976264378983,0.9046875746897421,0.006930976264378983,0.00013596775440067368
tfmmax_tw1,16.65690465603714,11.796620331132795,0.8693803364842014,11.75977131935039,0.9968763077264464,11.75977131935039,394.99021740557555
tfmmax_tw2,13.844677984260237,10.360091395134175,0.915862399267476,9.183877981132314,0.8864668882598571,9.183877981132314,261.01543432786775
tfmmax_frame,16.27367436030384,9.198502506096437,-0.18269815591915717,13.42460535101677,1.4594337874147911,13.42460535101677,621.8121617330642
Code/models/width_optimization/3W/rbf_models/per_output_models_B29_H45/it1/rbf_LOO_results.csv 0 → 100644
output,LOO_RMSE,LOO_MAE,LOO_R2,LOO_RMSE_STD,LOO_RMSE_CV,LOO_MAE_STD,LOO_SQERR_STD
exymax_tw1,0.009190462825822386,0.006555357477294431,0.8453857278703277,0.006441420285752449,0.98261922527694,0.006441420285752449,0.00017181266353825323
exymax_tw2,0.009561989511240931,0.00780680764044274,0.9010182687700412,0.0055213583363341355,0.7072491843824921,0.0055213583363341355,0.00010675387888540671
exymax_tw3,0.012254378933659786,0.008847059618380462,0.9382031312389641,0.008479347802664218,0.958436833074766,0.008479347802664218,0.0002587402983269403
tfmmax_tw1,15.60593832003407,9.246810431239657,0.897723967277053,12.571468000890956,1.3595464181269676,12.571468000890956,473.29935967123816
tfmmax_tw2,10.654495593789711,8.313696872079788,0.9515075314778572,6.663386577203453,0.8014950123550166,6.663386577203453,171.8915554159779
tfmmax_tw3,15.147861114739676,9.998707194184908,0.9605655290927848,11.37908391727547,1.138055520206989,11.37908391727547,410.83801514453637
tfmmax_frame,7.270920660929447,5.176765850734426,0.9558023431350221,5.105622644124867,0.9862572098756474,5.105622644124867,119.43968264760818
Code/models/width_optimization/3W/rbf_models/per_output_models_B34_H45/it1/rbf_LOO_results.csv 0 → 100644
output,LOO_RMSE,LOO_MAE,LOO_R2,LOO_RMSE_STD,LOO_RMSE_CV,LOO_MAE_STD,LOO_SQERR_STD
exymax_tw1,0.01271109892847045,0.008746127304020301,0.8089139553071438,0.009223735314460907,1.0546079417596705,0.009223735314460907,0.00032992539272167014
exymax_tw2,0.017087099155815577,0.01180245278726027,0.7023400693902224,0.012356013344326143,1.0469021623762302,0.012356013344326143,0.0005867726178967102
exymax_tw3,0.012853141734783946,0.009296509310458332,0.9438488063649062,0.008875706568775607,0.9547354036197938,0.008875706568775607,0.0002671563103569735
tfmmax_tw1,21.925609131135992,13.693767616424415,0.8780819688832557,17.123465310465633,1.2504568348251808,17.123465310465633,867.2586286515489
tfmmax_tw2,11.658665963212655,8.539220986273767,0.955307036290955,7.937644297230441,0.9295513384639745,7.937644297230441,232.8138565285835
tfmmax_tw3,17.170461532849675,11.112825367663884,0.9508512989207835,13.089303327483725,1.1778555762759486,13.089303327483725,569.9926878293359
tfmmax_frame,7.418295621994835,5.177950812310913,0.9442926024113952,5.312243906354122,1.025935567739253,5.312243906354122,129.21998015922588
Code/reports/width_optimization/2W/ml_optimization/it0/optimization_results_2W_B29_H30_TFD_W90.csv
...@@ -4,11 +4,11 @@ Configuration_B,29.0 ...@@ -4,11 +4,11 @@ Configuration_B,29.0
Configuration_H,30.0 Configuration_H,30.0
Configuration_TFD_W,90.0 Configuration_TFD_W,90.0
Iteration,0.0 Iteration,0.0
tw1_optimal,12.665567904044316 tw1_optimal,12.602604556600495
tw2_optimal,14.703450572353482 tw2_optimal,14.642223218431933
Objective_score,-1.689954500580596e-06 Objective_score,-1.703314522821794e-06
Exy_tw1,0.05013423656854821 Exy_tw1,0.05047722848321312
Exy_tw2,0.06146949165963539 Exy_tw2,0.06182536900480995
TFM_tw1,90.00000001387838 TFM_tw1,90.000000021604
TFM_tw2,90.00000002387473 TFM_tw2,90.00000001502372
TFM_frame,89.76113477601143 TFM_frame,89.6517557912324
Code/reports/width_optimization/2W/ml_optimization/it0/optimization_results_2W_B34_H30_TFD_W90.csv
...@@ -4,11 +4,11 @@ Configuration_B,34.0 ...@@ -4,11 +4,11 @@ Configuration_B,34.0
Configuration_H,30.0 Configuration_H,30.0
Configuration_TFD_W,90.0 Configuration_TFD_W,90.0
Iteration,0.0 Iteration,0.0
tw1_optimal,15.644808251437285 tw1_optimal,15.694454894049716
tw2_optimal,20.0 tw2_optimal,20.0
Objective_score,492.56724388692203 Objective_score,541.9474403032106
Exy_tw1,0.05093899889255882 Exy_tw1,0.05065454046456508
Exy_tw2,0.04702770620592467 Exy_tw2,0.04732158887659077
TFM_tw1,103.73892265716938 TFM_tw1,105.29400198067538
TFM_tw2,73.09918025112492 TFM_tw2,72.51250951714076
TFM_frame,96.59550068680014 TFM_frame,96.62331405344696
Code/reports/width_optimization/2W/rbf_optimization/it1/optimization_results_2W_B29_H30_TFD_W90.csv 0 → 100644
Parameter,Value
Configuration_W,2.0
Configuration_B,29.0
Configuration_H,30.0
Configuration_TFD_W,90.0
Iteration,1.0
tw1_optimal,12.684831662276643
tw2_optimal,14.905421089765571
Objective_score,-1.5458169860709181e-06
Exy_tw1,0.04531408098561937
Exy_tw2,0.06034192767159404
TFM_tw1,90.0000000200485
TFM_tw2,90.0000000392817
TFM_frame,87.5326351207742
Code/reports/width_optimization/2W/rbf_optimization/it1/optimization_results_2W_B34_H30_TFD_W90.csv 0 → 100644
Parameter,Value
Configuration_W,2.0
Configuration_B,34.0
Configuration_H,30.0
Configuration_TFD_W,90.0
Iteration,1.0
tw1_optimal,14.799811569069801
tw2_optimal,18.930435799107368
Objective_score,700.7361022316106
Exy_tw1,0.05458329763144634
Exy_tw2,0.049017179174410265
TFM_tw1,109.56930931180048
TFM_tw2,78.18164265531189
TFM_frame,96.73836958480248
Code/reports/width_optimization/3W/rbf_optimization/it1/optimization_results_3W_B29_H45_TFD_W90.csv 0 → 100644
Parameter,Value
Configuration_W,3.0
Configuration_B,29.0
Configuration_H,45.0
Configuration_TFD_W,90.0
Iteration,1.0
tw1_optimal,5.685399832978042
tw2_optimal,7.9673473591210415
tw3_optimal,9.016029941592292
Objective_score,-1.1929643499088454e-06
Exy_tw1,0.0467694639666424
Exy_tw2,0.048545210631678604
Exy_tw3,0.05550553027362139
TFM_tw1,90.00000001065337
TFM_tw2,90.00000001735125
TFM_tw3,90.00000001255168
TFM_frame,70.48753052656542
Code/reports/width_optimization/3W/rbf_optimization/it1/optimization_results_3W_B34_H45_TFD_W90.csv 0 → 100644
Parameter,Value
Configuration_W,3.0
Configuration_B,34.0
Configuration_H,45.0
Configuration_TFD_W,90.0
Iteration,1.0
tw1_optimal,6.81086626084703
tw2_optimal,9.021258980648522
tw3_optimal,9.654473348128397
Objective_score,-1.5352001974814094e-06
Exy_tw1,0.041897196711776205
Exy_tw2,0.04824066391668254
Exy_tw3,0.05385046859908042
TFM_tw1,90.00000001106143
TFM_tw2,90.0000000278992
TFM_tw3,90.00000001853854
TFM_frame,69.62930961050895
Code/src/width_optimization/ml_surrogate_train.py
...@@ -134,13 +134,13 @@ def get_adaptive_cv(n_samples: int): ...@@ -134,13 +134,13 @@ def get_adaptive_cv(n_samples: int):
Notes Notes
----- -----
- N <= 20: Leave-One-Out - N <= 20: Leave-One-Out
- 21 <= N <= 80: RepeatedKFold(5x5) - 21 <= N <= 80: RepeatedKFold(4x5)
- N > 80: KFold(5) - N > 80: KFold(5)
""" """
if n_samples <= 20: if n_samples <= 20:
return LeaveOneOut() return LeaveOneOut()
if n_samples <= 80: if n_samples <= 80:
return RepeatedKFold(n_splits=5, n_repeats=5, random_state=42) return RepeatedKFold(n_splits=4, n_repeats=5, random_state=42)
return KFold(n_splits=5, shuffle=True, random_state=42) return KFold(n_splits=5, shuffle=True, random_state=42)
...@@ -547,7 +547,7 @@ def main(): ...@@ -547,7 +547,7 @@ def main():
cv=cv, cv=cv,
n_jobs=-1, n_jobs=-1,
random_state=42, random_state=42,
optimizer_kwargs={"base_estimator": "RF"}, # optimizer_kwargs={"base_estimator": "RF"},
return_train_score=False, return_train_score=False,
verbose=0, verbose=0,
) )
......
Code/src/width_optimization/run_rbf.sh
#!/bin/bash #!/bin/bash
##python rbf_surrogate_train.py --W 2 --B 29 --it 0 python rbf_surrogate_train.py --W 2 --B 29 --it 1
##python rbf_optimization_de.py --W 2 --B 29 --TFD_W 90 --it 0 python rbf_optimization_de.py --W 2 --B 29 --TFD_W 90 --it 1
##python rbf_surrogate_train.py --W 2 --B 34 --it 0 python rbf_surrogate_train.py --W 2 --B 34 --it 1
##python rbf_optimization_de.py --W 2 --B 34 --TFD_W 90 --it 0 python rbf_optimization_de.py --W 2 --B 34 --TFD_W 90 --it 1
##python rbf_surrogate_train.py --W 3 --B 29 --it 0 python rbf_surrogate_train.py --W 3 --B 29 --it 1
##python rbf_optimization_de.py --W 3 --B 29 --TFD_W 90 --it 0 python rbf_optimization_de.py --W 3 --B 29 --TFD_W 90 --it 1
##python rbf_surrogate_train.py --W 3 --B 34 --it 0 python rbf_surrogate_train.py --W 3 --B 34 --it 1
##python rbf_optimization_de.py --W 3 --B 34 --TFD_W 90 --it 0 python rbf_optimization_de.py --W 3 --B 34 --TFD_W 90 --it 1
##python rbf_surrogate_train.py --W 5 --B 34 --it 2 ##python rbf_surrogate_train.py --W 5 --B 34 --it 2
##python rbf_optimization_de.py --W 5 --B 34 --TFD_W 90 --it 2 ##python rbf_optimization_de.py --W 5 --B 34 --TFD_W 90 --it 2
Manuscript/RESILINK_surrogate_optimization_manuscript.tex
...@@ -126,9 +126,11 @@ The admissible thickness ranges are defined according to the geometry family and ...@@ -126,9 +126,11 @@ The admissible thickness ranges are defined according to the geometry family and
\toprule \toprule
Family & Height $H$ & Width $B$ & Frame thickness & Design variables & Thickness bounds \\ Family & Height $H$ & Width $B$ & Frame thickness & Design variables & Thickness bounds \\
\midrule \midrule
2 windows & 30 cm & 29/34 cm & 30 mm & $t_{w,1},t_{w,2}$ & 10--20 mm \\ 2 windows B29 & 30 cm & 29 cm & 30 mm & $t_{w,1},t_{w,2}$ & 10--20 mm \\
3 windows & 45 cm & 29/34 cm & 30 mm & $t_{w,1},t_{w,2},t_{w,3}$ & 5--14 mm \\ 2 windows B34 & 30 cm & 34 cm & 30 mm & $t_{w,1},t_{w,2}$ & 10--20 mm \\
5 windows & 60 cm & 34 cm & 30 mm & $t_{w,1},\ldots,t_{w,5}$ & 5--12 mm \\ 3 windows B29 & 45 cm & 29 cm & 30 mm & $t_{w,1},t_{w,2},t_{w,3}$ & 5--14 mm \\
3 windows B34 & 45 cm & 34 cm & 30 mm & $t_{w,1},t_{w,2},t_{w,3}$ & 5--14 mm \\
5 windows B34 & 60 cm & 34 cm & 30 mm & $t_{w,1},\ldots,t_{w,5}$ & 5--12 mm \\
\bottomrule \bottomrule
\end{tabular} \end{tabular}
\end{table} \end{table}
...@@ -186,7 +188,7 @@ In addition, the maximum local shear distortion in each window is denoted by $\v ...@@ -186,7 +188,7 @@ In addition, the maximum local shear distortion in each window is denoted by $\v
\subsection{Design of experiments}\label{subsec:doe} \subsection{Design of experiments}\label{subsec:doe}
The FEM campaign is designed to cover the admissible design domain of each device family while ensuring a sufficiently homogeneous exploration of the multidimensional parameter space. The design variables correspond to the window thicknesses $t_{w,i}$, whose combinations are generated using a Design of Experiments (DoE) strategy based on Latin Hypercube Sampling (LHS) optimized with the maximin criterion \cite{Joseph2008}. This approach provides a near-random yet space-filling distribution of samples, reducing clustering effects and improving the representation of the admissible domain. The FEM campaign is designed to cover the admissible design domain of each device family while ensuring a homogeneous exploration of the multidimensional parameter space. The design variables correspond to the window thicknesses $t_{w,i}$, whose combinations are generated using a Design of Experiments (DoE) strategy based on Latin Hypercube Sampling (LHS) optimized with the maximin criterion \cite{Joseph2008}. This approach provides a near-random yet space-filling distribution of samples, reducing clustering effects and improving the representation of the admissible domain.
To improve the robustness of the surrogate models near the admissible limits, the sampling domain is extended slightly beyond the actual optimization ranges (see Table~\ref{tab:families}). This strategy reduces the risk of extrapolation when evaluating candidate solutions close to the true design limits. The ranges employed during the DoE are summarized in Table~\ref{tab:families_doe}. To improve the robustness of the surrogate models near the admissible limits, the sampling domain is extended slightly beyond the actual optimization ranges (see Table~\ref{tab:families}). This strategy reduces the risk of extrapolation when evaluating candidate solutions close to the true design limits. The ranges employed during the DoE are summarized in Table~\ref{tab:families_doe}.
...@@ -198,9 +200,11 @@ To improve the robustness of the surrogate models near the admissible limits, th ...@@ -198,9 +200,11 @@ To improve the robustness of the surrogate models near the admissible limits, th
\toprule \toprule
Family & Height $H$ & Width $B$ & Frame thickness & Design variables & Thickness bounds (DoE) \\ Family & Height $H$ & Width $B$ & Frame thickness & Design variables & Thickness bounds (DoE) \\
\midrule \midrule
2 windows & 30 cm & 29/34 cm & 30 mm & $t_{w,1},t_{w,2}$ & 8--22 mm \\ 2 windows B29 & 30 cm & 29 cm & 30 mm & $t_{w,1},t_{w,2}$ & 8--22 mm \\
3 windows & 45 cm & 29/34 cm & 30 mm & $t_{w,1},t_{w,2},t_{w,3}$ & 4--16 mm \\ 2 windows B34 & 30 cm & 34 cm & 30 mm & $t_{w,1},t_{w,2}$ & 8--22 mm \\
5 windows & 60 cm & 34 cm & 30 mm & $t_{w,1},\ldots,t_{w,5}$ & 4--14 mm \\ 3 windows B29 & 45 cm & 29 cm & 30 mm & $t_{w,1},t_{w,2},t_{w,3}$ & 4--16 mm \\
3 windows B34 & 45 cm & 34 cm & 30 mm & $t_{w,1},t_{w,2},t_{w,3}$ & 4--16 mm \\
5 windows B34 & 60 cm & 34 cm & 30 mm & $t_{w,1},\ldots,t_{w,5}$ & 4--14 mm \\
\bottomrule \bottomrule
\end{tabular} \end{tabular}
\end{table} \end{table}
...@@ -213,17 +217,21 @@ The number of samples in all cases is defined as a power of two. This choice fac ...@@ -213,17 +217,21 @@ The number of samples in all cases is defined as a power of two. This choice fac
\subsection{Supervised ML surrogate models}\label{subsec:ml_models} \subsection{Supervised ML surrogate models}\label{subsec:ml_models}
This work provides a systematic comparison of surrogate modelling techniques in terms of predictive accuracy, robustness, and computational efficiency within the context of geometry optimization of BDSL dampers. The considered supervised surrogate models include Random Forest (RF), Gradient Boosting Regression (GBR), XGBoost, Support Vector Regression (SVR), Multilayer Perceptron (MLP), and Gaussian Process Regression (GPR). This work compares the predictive accuracy and robustness of several surrogate modelling techniques of different nature for the estimation of structural response quantities related to damage and distortion. The supervised surrogate set includes tree-based methods, namely Random Forest (RF), Gradient Boosting Regression (GBR), and XGBoost; kernel-based approaches, including Support Vector Regression (SVR) and Gaussian Process Regression (GPR); and neural-network-based models represented by Multilayer Perceptron (MLP).
The objective is not the development of a new constitutive formulation, but the integration of FEM-calibrated damage and distortion indicators into a surrogate-assisted optimization framework capable of distinguishing between damage in the dissipative windows and in the surrounding frame. The proposed methodology also promotes balanced window activation and incorporates an adaptive FEM validation loop to verify the optimized geometries before acceptance. The objective is the integration of FEM-calibrated damage and distortion indicators into a surrogate-assisted optimization framework capable of evaluating the structural performance of different geometrical configurations. The proposed methodology seeks to minimize local damage in both the dissipative windows and the surrounding frame while promoting a balanced contribution of all windows to the energy dissipation process. Following optimization, an adaptive FEM validation loop is incorporated to verify the predicted optimal geometries before acceptance.
For each geometry family, a separate supervised regression model is trained for every target output variable. The input vector contains only the window thicknesses associated with the corresponding device configuration. The predicted outputs include the local distortion indicators $\varepsilon_{xy,i}$, the window damage indicators $\TFD_i$, and the frame damage indicator $\TFD_f$. For every output variable, all candidate algorithms are evaluated independently and the selected surrogate model is stored as a serialized \texttt{joblib} file. For each geometry family, an independent supervised regression model is trained for every target output variable. The input vector contains only the window thicknesses associated with the corresponding device configuration. The predicted outputs include the local distortion indicators $\varepsilon_{xy,i}$, the window damage indicators $\TFD_i$, and the frame damage indicator $\TFD_f$. Consequently, for each geometry family, a total of $2W+1$ surrogate models are trained, where $W$ denotes the number of windows: $W$ models associated with distortion indicators, $W$ models associated with window damage indicators, and one additional model corresponding to the frame damage indicator. For every target output, all candidate algorithms are evaluated independently, and the selected surrogate is stored as a serialized \texttt{joblib} file.
Hyperparameter optimization is performed through Bayesian search using 40 iterations. Model performance is evaluated using the root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination ($R^2$). Since the available dataset size varies between geometry families and adaptive iterations, the cross-validation strategy is selected automatically. Leave-One-Out validation is used for datasets with $N\leq20$, repeated five-fold cross-validation with five repetitions is adopted for $21\leq N\leq80$, and standard shuffled five-fold cross-validation is employed for larger datasets. For small datasets, the hyperparameter search spaces of tree-based models are reduced to mitigate overfitting. For each geometry family and target output, hyperparameter optimization is performed independently for every ML model using Bayesian optimization. This iterative search strategy constructs a probabilistic surrogate of the objective function, typically based on Gaussian processes as adopted in the present study, to guide the exploration of the hyperparameter space \cite{Snoek2012}. Compared with conventional grid-search strategies, Bayesian optimization is significantly more efficient because it concentrates the search on regions that are more likely to improve model performance while avoiding less promising combinations.
The surrogate selection process is not based exclusively on the minimum average RMSE. Models whose RMSE falls within a 5\% tolerance band of the best-performing candidate are considered competitive. When fold-level dispersion metrics are available for all competitive candidates, the model with the lowest RMSE coefficient of variation is selected. This criterion favours not only accurate but also stable surrogates, which is particularly important in optimization problems where small local prediction errors near active damage constraints may alter the final accepted geometry. A total of 40 optimization iterations are performed for each ML model. The first 10 iterations are devoted to random exploration of the hyperparameter space, whereas the remaining 30 iterations are guided by the probabilistic surrogate model. This strategy considerably reduces the number of hyperparameter evaluations required compared with exhaustive grid-search procedures, particularly considering the large number of geometry families, target outputs, and candidate surrogate models analysed in this work. The optimal hyperparameter configuration is selected as the one minimizing the mean absolute error (MAE) during cross-validation.
\begin{figure}[t] Model performance is evaluated primarily through the root mean squared error (RMSE). Since the available dataset size varies between geometry families and adaptive iterations, the cross-validation strategy is selected automatically. Leave-One-Out validation is employed for datasets with $N\leq20$, repeated five-fold cross-validation with five repetitions is adopted for $21\leq N\leq80$, and standard shuffled five-fold cross-validation is used for larger datasets. For small datasets, the hyperparameter search spaces of tree-based models are additionally restricted to mitigate overfitting.
The surrogate selection process is not based exclusively on the minimum average RMSE. Models whose RMSE falls within a 5\% tolerance band of the best-performing candidate are considered competitive. When fold-level dispersion metrics are available for all competitive candidates, the model with the lowest coefficient of variation of the RMSE is selected. This criterion favours not only accurate but also stable surrogates, which is particularly important in optimization problems where small local prediction errors near active damage constraints may alter the final accepted geometry.
\begin{figure}[ht!]
\centering \centering
\fbox{\parbox[c][0.24\textheight][c]{0.85\textwidth}{\centering Placeholder for supervised surrogate workflow: FEM dataset generation, adaptive cross-validation, Bayesian hyperparameter optimization, model selection, and surrogate persistence.}} \fbox{\parbox[c][0.24\textheight][c]{0.85\textwidth}{\centering Placeholder for supervised surrogate workflow: FEM dataset generation, adaptive cross-validation, Bayesian hyperparameter optimization, model selection, and surrogate persistence.}}
\caption{Workflow of the supervised surrogate training strategy. A separate surrogate model is trained for each output variable, and the selected algorithm may differ depending on the predicted quantity.} \caption{Workflow of the supervised surrogate training strategy. A separate surrogate model is trained for each output variable, and the selected algorithm may differ depending on the predicted quantity.}
......
Manuscript/wileyNJD-AMA.bib
...@@ -584,8 +584,23 @@ steel from coupon test results available. First, the theory of metal plasticity ...@@ -584,8 +584,23 @@ steel from coupon test results available. First, the theory of metal plasticity
doi = {10.1016/j.envsoft.2017.03.010}, doi = {10.1016/j.envsoft.2017.03.010},
keywords = {Design of computer experiments, Monte Carlo simulation, Optimal Latin hypercube sampling, Sensitivity analysis, Sequential sampling, Uncertainty analysis}, keywords = {Design of computer experiments, Monte Carlo simulation, Optimal Latin hypercube sampling, Sensitivity analysis, Sequential sampling, Uncertainty analysis},
shorttitle = {Progressive {Latin} {Hypercube} {Sampling}}, shorttitle = {Progressive {Latin} {Hypercube} {Sampling}},
url = {https://www.sciencedirect.com/science/article/pii/S1364815216305096},
urldate = {2026-05-06}, urldate = {2026-05-06},
} }
@InProceedings{Snoek2012,
author = {Snoek, Jasper and Larochelle, Hugo and Adams, Ryan P.},
booktitle = {Proceedings of the 26th {International} {Conference} on {Neural} {Information} {Processing} {Systems} - {Volume} 2},
title = {Practical {Bayesian} optimization of machine learning algorithms},
year = {2012},
address = {Red Hook, NY, USA},
month = dec,
pages = {2951--2959},
publisher = {Curran Associates Inc.},
series = {{NIPS}'12},
volume = {2},
abstract = {The use of machine learning algorithms frequently involves careful tuning of learning parameters and model hyperparameters. Unfortunately, this tuning is often a "black art" requiring expert experience, rules of thumb, or sometimes brute-force search. There is therefore great appeal for automatic approaches that can optimize the performance of any given learning algorithm to the problem at hand. In this work, we consider this problem through the framework of Bayesian optimization, in which a learning algorithm's generalization performance is modeled as a sample from a Gaussian process (GP). We show that certain choices for the nature of the GP, such as the type of kernel and the treatment of its hyperparameters, can play a crucial role in obtaining a good optimizer that can achieve expertlevel performance. We describe new algorithms that take into account the variable cost (duration) of learning algorithm experiments and that can leverage the presence of multiple cores for parallel experimentation. We show that these proposed algorithms improve on previous automatic procedures and can reach or surpass human expert-level optimization for many algorithms including latent Dirichlet allocation, structured SVMs and convolutional neural networks.},
doi = {10.5555/2999325.2999464},
urldate = {2026-04-21},
}
@Comment{jabref-meta: databaseType:bibtex;} @Comment{jabref-meta: databaseType:bibtex;}
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment