inputMLPhotovoltaics.inp

#######################################
### File containing input options for MLPhotovoltaics.py
### The name of this file can be changed, but then also need to update 'input_file_name' in ML_d_a.py
### Lines that start with '#' are ignored. All comments after '#' are ignored. Empty lines are ignored
### Options can be in any order
#######################################
#######################################
########## Parallelization ############
#######################################
NCPU = 1                              # select number of CPUs (-1 means all CPUs in a node) (ONLY USED IF "optimize_hyperparams = True")
#######################################
############ Verbose options ##########
#######################################
print_log = True                      # choose whether information is also written into a log file (Default: True)
log_name  = 'log_MLPhotovoltaics.log' # name of log file
print_progress_every_x_percent =  10  # print progress in standard output at different interval
#######################################
######### Data base options ###########
#######################################
## Any number of electronic (physical) descriptors can be specified, classified in any number of groups. Use notation: xcols_elec0, xcols_elec1 etc.
## Each of these groups can contain any number of descriptors, but the total number of groups should be the same as
## the length of 'gamma_el' and 'gamma_el_lim', specified in 'Machine Learning Algorithm options'. Each of these gammas
## controls the weight of each group of descriptors
#######################################
db_file = 'database.csv'              # name of input file with database
Ndata =  566                          # number of d/a pairs
xcols_struc = ["SMILES-DFP","SMILES-AFP"]                    # name of descriptors that will be affected by gamma_d and gamma_a, respectively
xcols_elec0 = ["HOMO-D","LUMO-D","LUMO-A","Reor-D","Reor-A"] # name of descriptors affected by gamma_el[0]
xcols_elec1 = ["sum of f-D", "sum of f-A"]                   # name of descriptors affected by gamma_el[1]
xcols_elec2 = ["XLOGP3-D","XLOGP3-A"]                        # name of descriptors affected by gamma_el[2]
ycols = [ "PCE" ]                     # name of target property
FP_length = 2048						     # number of entries of each fingerprint
#######################################
####### Learning Curve options ########
#######################################
lc_CV = 'kf'								  # select 'loo' or 'kf'
lc_kfold = 5 								  # only used if lc_CV='kf'
plot_lc = None		  						  # decide filename for learning curve graph. If None, it is not plotted
#######################################
####### Output prediction csv #########
#######################################
prediction_csv_file_name = 'pred.csv' # Select None if no output prediction csv is wanted
columns_labels_prediction_csv = []	  # Select extra labels to print in output prediction csv
#######################################
# Machine Learning Algorithm options ##
#######################################
ML = 'kNN'                            # 'kNN' or 'KRR' or 'SVR'
CV = 'loo'                            # select 'loo' or 'kf' or 'last'
kfold = 3                             # only used if CV='kf'
Nlast = 10                            # only used if CV='last'
plot_target_predictions = "pred.png"  # decide filename for graph plotting predicted and actual values of target property. If None, it is not plotted
### General hyperparameters ###########
optimize_hyperparams = False          # whether hyperparameters are optimized (True) or just use initial values (False). If gamma_X=0.0, then that hyperparam is not optimized anyway
gamma_el = [0.4792,0.0512,0.1104]     # hyperparameters with weights of d_el
gamma_d  = 5.5437            	        # hyperparameter with weight of d_fp_d
gamma_a  = 1.8608            	        # hyperparameter with weight of d_fp_a
gamma_el_lim = [(0.0, 2.0),(0.0,2.0),(0.0,2.0)] # range in which each gamma_el is optimized
gamma_d_lim  = (0.0, 6.0)             # range in which gamma_d is optimized
gamma_a_lim  = (0.0, 6.0)             # range in which gamma_a is optimized
weight_RMSE = 'linear'                # select one: 'linear', 'PCE', 'PCE2'. Default: 'linear'
### k-Nearest Neighbors ('kNN') #######
Neighbors = 3                         # number of nearest-neighbors
plot_kNN_distances = None 	           # decide filename for graph of kNN-distances vs rmse at each CV step is printed. If None, it is not plotted
### Kernel Ridge Regression ('KRR') ###
alpha = 0.3                           # kernel hyperparameter
alpha_lim  = (0.0, 2.0)               # range in which alpha hyperparam is optimized
### Support Vector Regression ('SVR')##
C = 1.0                               # regularization hyperparameter
epsilon = 0.1                         # epsilon-SVR
C_lim = (0.01,10)                     # range in which C is optimized
epsilon_lim = (0.001,10)              # range in which epsilon is optimized
#######################################