def time_euclidean_distance_learning():
"run learning with Euc distance to see the time taken and compare with chemical_distance"
db = DB("y4_python/11k_molecule_database_eV.db")
reg = MyRegression(db)
mol_list = db.get_mol_ids()
pm7_energies = db.get_pm7_energies()
blyp_energies = db.get_blyp_energies()
deviation_list = (list(
map(reg.distance_from_regress, pm7_energies, blyp_energies)))
main_euclidean_distance(k_neighbours=5,
k_folds=10,
metric_params={},
mol_list=mol_list,
deviation_list=deviation_list)
def time_chemical_distance_learning():
"run learning with different params to see what takes the longest. Use limited set of data so it doesn't take so long on each run."
db = DB("y4_python/11k_molecule_database_eV.db")
reg = MyRegression(db)
cutoff = 1000
mol_list = db.get_mol_ids()[:cutoff]
pm7_energies = db.get_pm7_energies()[:cutoff]
blyp_energies = db.get_blyp_energies()[:cutoff]
deviation_list = (list(
map(reg.distance_from_regress, pm7_energies, blyp_energies)))
fingerprint_list = db.get_fingerprints()[:cutoff]
homo_list = db.get_homo_molecular_orbitals()[:cutoff]
lumo_list = db.get_lumo_molecular_orbitals()[:cutoff]
params = MetricParams(homo_coeff=1.0,
lumo_coeff=0.0,
fingerprint_list=fingerprint_list,
homo_orbital_list=homo_list,
lumo_orbital_list=lumo_list,
c_struct=0.0,
c_orbital=1.0,
inertia_coefficient=0.0,
IPR_coefficient=0.0,
N_coefficient=0.0,
O_coefficient=0.0,
S_coefficient=0.0,
P_coefficient=0.0,
radial_distribution_coeff=1.0)
results = main_chemical_distance(k_neighbours=5,
k_folds=10,
metric_params=params,
mol_list=mol_list,
deviation_list=deviation_list,
save=False)
print(results)
if __name__ == "__main__":
c_struct_lim = (0.0, 10.0)
c_orbital_lim = (0.0, 10.0)
bounds = [c_struct_lim] + [
c_orbital_lim
] # this will be whatever hyperparameters you want to optimize. The length of this sequence is what indicates the number of hyperparameters mini_args = (X, y, condition,fixed_hyperparams) # this will be a tuple with the rest of arguments needed for your function NCPU = 20 # this will be the number of CPUs you want
num_CPU = 20
db = DB("y4_python/11k_molecule_database_eV.db")
my_regression = MyRegression(db)
mol_list = db.get_mol_ids()
pm7_energies = db.get_pm7_energies()
blyp_energies = db.get_blyp_energies()
deviation_list = (list(
map(my_regression.distance_from_regress, pm7_energies, blyp_energies)))
fingerprint_list = db.get_fingerprints()
homo_molecular_orbital_list = db.get_homo_molecular_orbitals()
lumo_molecular_orbital_list = db.get_lumo_molecular_orbitals()
fixed_hyperparams = MetricParams(
homo_coeff=1.0,
lumo_coeff=0.0,
fingerprint_list=fingerprint_list,
homo_orbital_list=homo_molecular_orbital_list,
lumo_orbital_list=lumo_molecular_orbital_list,
c_struct=0.0 # TO BE OPTIMISED...
,
c_orbital=0.0 # TO BE OPTIMISED...