def eval_function(self, target, data, function, dt, i):
"""
function that sends target and simulated volts to scorefunctions.py so they
can be evaluated, then adds Kyung AP penalty function at the end
Parameters
-------------------------------------------------------------
target: target volt for this stim
data: set of volts with shape (nindvs, ntimsteps)
function: string containing score function to use
dt: dt of the stim as it is a parameter for some sfs
i: index of the stim
Returns
------------------------------------------------------------
score: scores for each individual in an array corresponding to this score function
with shape (nindv, 1)
"""
num_indvs = data.shape[0]
# for indv in range(num_indvs):
# print(np.isnan(data[indv,:]), "is Nan for ", indv)
# print(1/0)
if function in custom_score_functions:
score = [
getattr(sf, function)(target, data[indv, :], dt)
for indv in range(num_indvs)
]
else:
score = sf.eval_efel(function, target, data, dt)
return score + check_ap_at_zero(i, data) # here is I am adding penalty
def eval_function(self,target, data, function, dt,i):
"""
function that sends target and simulated volts to scorefunctions.py so they
can be evaluated, then adds Kyung AP penalty function at the end
Parameters
-------------------------------------------------------------
target: target volt for this stim
data: set of volts with shape (nindvs, ntimsteps)
function: string containing score function to use
dt: dt of the stim as it is a parameter for some sfs
i: index of the stim
Returns
------------------------------------------------------------
score: scores for each individual in an array corresponding to this score function
with shape (nindv, 1)
"""
#num_indvs = data.shape[0]
# for indv in range(num_indvs):
# print(np.isnan(data[indv,:]), "is Nan for ", indv)
# print(1/0)
score = []
if function in custom_score_functions:
with open('/tmp/Data/VHotP' + str(i) + '.dat', 'rb') as f:
for voltset in read_in_chunks(f,i): # use this for customsfs and for setting up efel
score.append(getattr(sf, function)(target, voltset, dt)) # stream data here
else:
score = sf.eval_efel(function, target, i, dt)
return np.array(score) + check_ap_at_zero(i, self.nindv)# here is I am adding penalty
def eval_function(target, data, function, dt):
'''changed from hoc eval so that it returns eval for list of indvs, not just one'''
if function in custom_score_functions:
score = [getattr(sf, function)(target, data[i], dt) for indv in data]
else:
score = sf.eval_efel(function, target, data, dt)
score = np.reshape(score, self.nindv)
return score
def eval_stim_sf_pair(self,perm):
"""
function that evaluates a stim and score function pair on line 252. Sets i as stim # and sets j as
score function #. Evaluates volts for that score function in efel or custom. Normalize scores
then SENT BACK to MAPPER.
Arguments
--------------------------------------------------------------------
perm: pair of ints where first is the stim and second is the score function label index
to run
Returns
---------------------------------------------------------------------
scores: normalized+weighted scores with the shape (nindv, 1), and sends them back to map
to be stacked then summed.
"""
i = perm[0]
j = perm[1]
counter = 0
curr_data_volt = self.getVolts(i)#[:,:]
curr_target_volt = self.target_volts_list[i]
curr_sf = score_function_ordered_list[j].decode('ascii')
curr_weight = self.weights[len(score_function_ordered_list)*i + j]
transformation = h5py.File(scores_path+self.opt_stim_list[i]+'_scores.hdf5', 'r')['transformation_const_'+curr_sf][:]
if curr_weight == 0:
print("BAD WEIGHTS")
curr_scores = np.zeros(self.nindv)
else:
num_indvs = curr_data_volt.shape[0]
if curr_sf in custom_score_functions:
score = [getattr(sf, curr_sf)(curr_target_volt, curr_data_volt[indv,:], self.dts[i]) for indv in range(num_indvs)]
else:
score = sf.eval_efel(curr_sf, curr_target_volt, curr_data_volt, self.dts[i])
curr_scores = score
norm_scores = self.normalize_scores(curr_scores, transformation, i)
for k in range(len(norm_scores)):
if np.isnan(norm_scores[k]):
norm_scores[k] = 1
return norm_scores * curr_weight
def eval_function(target, data, function, dt):
if function in custom_score_functions:
score = getattr(sf, function)(target, data, dt)
else:
score = sf.eval_efel(function, target, data, dt)
return score