def _get_fitness(anion, genes, target, models, deslists):
"""
the fitness function passed to the engine.
Parameters
----------
anion : RDKit Mol Object
the anion comprising the IL
genes : str
the smiles string representing the cation of the IL
target : list, int, or array
the target property values of the IL
models : array of, or single Keras model
array or single keras model to use in the prediction
of the targets
deslists : array of, or single pandas dataFrame
contains the mean and stds of the model inputs
"""
predictions = []
for i, name in enumerate(models):
cation = Chem.MolFromSmiles(genes)
model = name
deslist = deslists[i]
if isinstance(deslist, list):
deslist = deslist[0]
feature_vector = []
for item in deslist:
if "anion" in item:
with genetic.suppress_rdkit_sanity():
feature_vector.append(
calculator([item.partition('-')[0]
]).CalcDescriptors(anion)[0])
elif "cation" in item:
with genetic.suppress_rdkit_sanity():
feature_vector.append(
calculator([item.partition('-')[0]
]).CalcDescriptors(cation)[0])
elif "Temperature, K" in item:
feature_vector.append(298.15)
elif "Pressure, kPa" in item:
feature_vector.append(101.325)
else:
print("unknown descriptor in list: %s" % item)
features_normalized = (feature_vector - deslist.iloc[0].values) /\
deslist.iloc[1].values
prediction = np.round(np.exp(
model.predict(np.array(features_normalized).reshape(1, -1))[0]),
decimals=2)
predictions.append(prediction[0])
predictions = np.array(predictions)
error = abs((predictions - target) / target)
error = np.average(error)
return 1 - error, predictions
def get_fitness(anion, genes, target):
model_ID = "density"
cation = Chem.MolFromSmiles(genes)
model = genetic.load_data("{}_qspr.h5".format(model_ID), h5File=True)
deslist = genetic.load_data("{}_desc.csv".format(model_ID))
feature_vector = []
with genetic.suppress_rdkit_sanity():
for item in deslist:
if "anion" in item:
feature_vector.append(
calculator([item.partition('-')[0]
]).CalcDescriptors(anion)[0])
elif "cation" in item:
feature_vector.append(
calculator([item.partition('-')[0]
]).CalcDescriptors(cation)[0])
elif "Temperature, K" in item:
feature_vector.append(298.15)
elif "Pressure, kPa" in item:
feature_vector.append(101.325)
else:
print("unknown descriptor in list: %s" % item)
features_normalized = (feature_vector - deslist.iloc[0].values) /\
deslist.iloc[1].values
prediction = exp(
model.predict(np.array(features_normalized).reshape(1, -1))[0])
error = abs((prediction - target) / target)
return 1 - error, prediction
def _get_fitness(anion, genes, target, models, deslists):
"""
the fitness function passed to the engine. In this case fitness
is determined by a model developed by the salty module.
The fitness function can handle multi-output models
"""
predictions = []
for i, name in enumerate(models):
cation = Chem.MolFromSmiles(genes)
model = name
deslist = deslists[i]
if isinstance(deslist, list):
deslist = deslist[0]
feature_vector = []
for item in deslist:
if "anion" in item:
with genetic.suppress_rdkit_sanity():
feature_vector.append(
calculator([item.partition('-')[0]
]).CalcDescriptors(anion)[0])
elif "cation" in item:
with genetic.suppress_rdkit_sanity():
feature_vector.append(
calculator([item.partition('-')[0]
]).CalcDescriptors(cation)[0])
elif "Temperature, K" in item:
feature_vector.append(298.15)
elif "Pressure, kPa" in item:
feature_vector.append(101.325)
else:
print("unknown descriptor in list: %s" % item)
features_normalized = (feature_vector - deslist.iloc[0].values) /\
deslist.iloc[1].values
prediction = np.round(np.exp(
model.predict(np.array(features_normalized).reshape(1, -1))[0]),
decimals=2)
predictions.append(prediction[0])
predictions = np.array(predictions)
error = abs((predictions - target) / target)
error = np.average(error)
return 1 - error, predictions
def _get_fitness(anion, genes, target, model_ID):
"""
the fitness function passed to the engine. In this case fitness
is determined by a model developed by the salty module. It is
important to note that the fitness function can handle multi-
output models
"""
cation = Chem.MolFromSmiles(genes)
model = genetic.load_data("{}.sav".format(model_ID), pickleFile=True)
deslist = genetic.load_data("{}_descriptors.csv".format(model_ID))
feature_vector = []
for item in deslist:
if "anion" in item:
with genetic.suppress_rdkit_sanity():
feature_vector.append(
calculator([item.partition('-')[0]
]).CalcDescriptors(anion)[0])
elif "cation" in item:
with genetic.suppress_rdkit_sanity():
feature_vector.append(
calculator([item.partition('-')[0]
]).CalcDescriptors(cation)[0])
elif "Temperature, K" in item:
feature_vector.append(298.15)
elif "Pressure, kPa" in item:
feature_vector.append(101.325)
else:
print("unknown descriptor in list: %s" % item)
features_normalized = (feature_vector - deslist.iloc[0].values) /\
deslist.iloc[1].values
prediction = np.round(np.exp(
model.predict(np.array(features_normalized).reshape(1, -1))[0]),
decimals=2)
error = abs((prediction - target) / target)
error = np.average(error)
return 1 - error, prediction