1. FS_3: results from 1st version of the code

2. FS_4: results from 2nd version of the code

3. FS_5: results from 3rd version of the code: (3 different challenges: ICaL=8, ICaL+8, ICaL*8 -> but the results are the same between all the tests in these 3 different challenges)

   a) COST1: using of cost function_1 (mean squared differences) -> every time use EAD=100000 in order to be sure to give importance to EADs and to be able to get rid of them during the generations

   b) COST2: using of cost function_2 (check the boundaries) -> inside cost2 sometimes i used EAD=1000 and other times i used EAD=100000

   c) name of the files: 1st number indicate number of individuals while the second one indicate the number of generations used

"FS_data": excel file that will fill with the values of the conductances

"FS_errors": excel file that will fill with the errors

"EADs": excel file that will fill with the EADs

"important functions": functions used for the EADs (run_EAD and detect_EAD#

"tor_ord_endo" and "tor_ord_enndo2": models

"supercell_ga": just to try with the ideal cases or tests

1. preprocessing and proto schedule
2. old feature targets: {'dvdt_max': [80, 86, 92], 'apd10': [5, 15, 30], 'apd50': [200, 220, 250], 'apd90': [250, 270, 300], 'cat_amp': [2.8E-4, 3.12E-4, 4E-4], 'cat10': [80, 100, 120], 'cat50': [200, 220, 240], 'cat90': [450, 470, 490]}

1. just 2 tunable parameters (Ikr and ICaL) instead of 5 tunable parameters
2. error goes to 0
3. the new feature targets: {'Vm_peak': [10, 33, 55], 'dvdt_max': [100, 347, 1000], 'apd40': [85, 198, 320], 'apd50': [110, 220, 430], 'apd90': [180, 271, 440], 'triangulation': [50, 73, 150], 'RMP': [-95, -88, -80], 'cat_amp': [3E-41e5, 3.12E-41e5, 8E-4*1e5], 'cat_peak': [40, 58, 60], 'cat90': [350, 467, 500]}

1. little error 1e-5 (inside "get_feature_error" line " ap_features['cat_amp'] = cat_amp*1e-5" )
2. computation of EADs whith run_EAD and detect_EAD

1. use "concat" instead of "append"
2. adjustments on return 500000 condition
3. excel file EADs
4. Graph EADs vs Generations

1. just consider ICaL and IKr as blockers -> changed the interval of lower and upper tunable parameters from 0.1 to 1 instead of 0.1 to 10

KRISTIN ADDING NEW FUNCTIONS ABOUT THE ERROR COMPUTATION:

1. def closest
2. def check_physio_torord
3. def get_torord_phys_data
4. Modify computation of fitness error inside get_feature_error