Update tutorial artifacts for braindecode/plot_data_augmentation_search

README.md

@@ -6,5 +6,5 @@ These files are meant to be loaded by the tutorial so the docs can plot the offl
 
 ## Stored files
 
-- `search_results.csv`: serialized `GridSearchCV.cv_results_`
+- `search_results.csv`: tidy cross-validation search summary
 - `metadata.json`: summary metrics for the saved search

metadata.json

@@ -1,15 +1,21 @@
 {
-  "best_augmentation": "SmoothTimeMask()",
+  "best_augmentation": "SmoothTimeMask",
+  "best_candidate": "SmoothTimeMask(mask_len_samples=300)",
+  "best_magnitude": 300.0,
+  "best_relative_validation_improvement": 0.6438356164383561,
   "chance_level": 0.25,
   "cv_splits": 2,
   "display_metric_key": "eval_accuracy",
   "display_metric_name": "accuracy",
   "display_split_name": "held-out session",
-  "epochs_requested": 20,
-  "eval_accuracy": 0.6145833333333334,
-  "search_candidates": 6,
+  "epochs_requested": 100,
+  "eval_accuracy": 0.6805555555555556,
+  "identity_validation_score": 0.2534722222222222,
+  "patience": 20,
+  "search_candidates": 16,
+  "search_magnitudes_per_augmentation": 5,
   "short_run_epochs": 2,
-  "training_score": 0.7048611111111112,
+  "training_score": 0.8020833333333333,
   "tutorial": "plot_data_augmentation_search",
-  "validation_score": 0.3993055555555556
+  "validation_score": 0.41666666666666663
 }

search_results.csv

@@ -1,7 +1,17 @@
-mean_fit_time,std_fit_time,mean_score_time,std_score_time,param_iterator_train__transforms,params,split0_test_score,split1_test_score,mean_test_score,std_test_score,rank_test_score,split0_train_score,split1_train_score,mean_train_score,std_train_score
-19.13320744037628,0.08719742298126221,0.3246210813522339,0.0034400224685668945,FTSurrogate(),{'iterator_train__transforms': FTSurrogate()},0.4166666666666667,0.375,0.39583333333333337,0.020833333333333343,2,0.7638888888888888,0.6666666666666666,0.7152777777777777,0.048611111111111105
-19.76908302307129,0.1334381103515625,0.25862085819244385,0.005473017692565918,FTSurrogate(),{'iterator_train__transforms': FTSurrogate()},0.4513888888888889,0.3333333333333333,0.3923611111111111,0.05902777777777779,3,0.6388888888888888,0.5069444444444444,0.5729166666666666,0.06597222222222221
-19.327903866767883,0.15474402904510498,0.36883795261383057,0.10311996936798096,SmoothTimeMask(),{'iterator_train__transforms': SmoothTimeMask()},0.4305555555555556,0.3680555555555556,0.3993055555555556,0.03125,1,0.7430555555555556,0.6666666666666666,0.7048611111111112,0.038194444444444475
-18.767938494682312,0.08156955242156982,0.28813016414642334,0.010505080223083496,SmoothTimeMask(),{'iterator_train__transforms': SmoothTimeMask()},0.3611111111111111,0.3819444444444444,0.3715277777777778,0.010416666666666657,5,0.5902777777777778,0.5763888888888888,0.5833333333333333,0.006944444444444475
-19.036476135253906,0.13442707061767578,0.30655086040496826,0.03131401538848877,ChannelsDropout(),{'iterator_train__transforms': ChannelsDropout()},0.4375,0.3125,0.375,0.0625,4,0.7638888888888888,0.5763888888888888,0.6701388888888888,0.09375
-18.096240043640137,0.7642838954925537,0.31714797019958496,0.014190912246704102,ChannelsDropout(),{'iterator_train__transforms': ChannelsDropout()},0.2847222222222222,0.2708333333333333,0.2777777777777778,0.0069444444444444475,6,0.4027777777777778,0.4375,0.4201388888888889,0.017361111111111105
+candidate_label,augmentation,magnitude,display_magnitude,axis_label,sort_order,mean_training_accuracy,std_training_accuracy,mean_validation_accuracy,std_validation_accuracy,rank_validation_accuracy,relative_validation_improvement,relative_training_improvement,relative_validation_improvement_pct,relative_training_improvement_pct
+IdentityTransform(),IdentityTransform,0.0,0.0,Identity baseline,0,0.26041666666666663,0.03819444444444445,0.2534722222222222,0.00347222222222221,9,0.0,0.0,0.0,0.0
+FTSurrogate(phase_noise_magnitude=0.1),FTSurrogate,0.1,0.1,Phase noise magnitude,1,0.24652777777777776,0.024305555555555552,0.21875,0.010416666666666657,15,-0.136986301369863,-0.05333333333333323,-13.698630136986301,-5.333333333333323
+FTSurrogate(phase_noise_magnitude=0.3),FTSurrogate,0.3,0.3,Phase noise magnitude,1,0.2638888888888889,0.02083333333333333,0.2361111111111111,0.013888888888888895,12,-0.06849315068493145,0.01333333333333342,-6.849315068493144,1.333333333333342
+FTSurrogate(phase_noise_magnitude=0.5),FTSurrogate,0.5,0.5,Phase noise magnitude,1,0.2743055555555555,0.04513888888888888,0.2361111111111111,0.03472222222222221,12,-0.06849315068493145,0.053333333333333455,-6.849315068493144,5.3333333333333455
+FTSurrogate(phase_noise_magnitude=0.7),FTSurrogate,0.7,0.7,Phase noise magnitude,1,0.2326388888888889,0.02430555555555554,0.22569444444444442,0.0034722222222222238,14,-0.10958904109589052,-0.10666666666666647,-10.958904109589051,-10.666666666666647
+FTSurrogate(phase_noise_magnitude=0.9),FTSurrogate,0.9,0.9,Phase noise magnitude,1,0.22916666666666669,0.02083333333333333,0.21875,0.00347222222222221,15,-0.136986301369863,-0.11999999999999977,-13.698630136986301,-11.999999999999977
+SmoothTimeMask(mask_len_samples=100),SmoothTimeMask,100.0,0.4,Mask length (s),2,0.2569444444444444,0.0,0.2673611111111111,0.010416666666666685,8,0.0547945205479452,-0.013333333333333308,5.47945205479452,-1.3333333333333308
+SmoothTimeMask(mask_len_samples=200),SmoothTimeMask,200.0,0.8,Mask length (s),2,0.6458333333333333,0.18055555555555552,0.3784722222222222,0.13541666666666666,4,0.4931506849315068,1.48,49.31506849315068,148.0
+SmoothTimeMask(mask_len_samples=300),SmoothTimeMask,300.0,1.2,Mask length (s),2,0.8020833333333333,0.024305555555555525,0.41666666666666663,0.06944444444444445,1,0.6438356164383561,2.08,64.3835616438356,208.0
+SmoothTimeMask(mask_len_samples=400),SmoothTimeMask,400.0,1.6,Mask length (s),2,0.7534722222222223,0.017361111111111105,0.38541666666666663,0.03819444444444445,3,0.5205479452054793,1.893333333333334,52.05479452054793,189.3333333333334
+SmoothTimeMask(mask_len_samples=500),SmoothTimeMask,500.0,2.0,Mask length (s),2,0.7673611111111112,0.017361111111111105,0.3993055555555556,0.024305555555555552,2,0.5753424657534247,1.9466666666666672,57.53424657534247,194.6666666666667
+ChannelsDropout(p_drop=0.2),ChannelsDropout,0.2,0.2,Drop probability,3,0.2361111111111111,0.0069444444444444475,0.24305555555555555,0.027777777777777762,11,-0.041095890410958846,-0.09333333333333327,-4.109589041095885,-9.333333333333327
+ChannelsDropout(p_drop=0.4),ChannelsDropout,0.4,0.4,Drop probability,3,0.5625,0.3194444444444444,0.3298611111111111,0.12152777777777778,6,0.3013698630136987,1.1600000000000001,30.136986301369873,116.00000000000001
+ChannelsDropout(p_drop=0.6),ChannelsDropout,0.6,0.6,Drop probability,3,0.27083333333333337,0.020833333333333343,0.24652777777777776,0.010416666666666657,10,-0.0273972602739726,0.04000000000000026,-2.73972602739726,4.000000000000026
+ChannelsDropout(p_drop=0.8),ChannelsDropout,0.8,0.8,Drop probability,3,0.6284722222222222,0.23263888888888892,0.3298611111111111,0.05902777777777779,6,0.3013698630136987,1.4133333333333336,30.136986301369873,141.33333333333334
+ChannelsDropout(p_drop=1.0),ChannelsDropout,1.0,1.0,Drop probability,3,0.5868055555555556,0.045138888888888895,0.3368055555555556,0.05208333333333334,5,0.32876712328767144,1.2533333333333339,32.87671232876714,125.33333333333339