def __init__(self):
"""Constructor"""
data = nrnUtils.readParamsCSV(paramsCSV)
super(hoc_evaluator, self).__init__()
self.orig_params = orig_params
self.opt_ind = params_opt_ind
data = np.array([data[i] for i in self.opt_ind])
self.orig_params = orig_params
self.pmin = np.array((data[:, 1]), dtype=np.float64)
self.pmax = np.array((data[:, 2]), dtype=np.float64)
self.pmax = np.delete(
self.pmax, 1, 0
) # need to delete second parameter because it is negative and BPOP cannot take negative params, we will add it back to every set of params in eval_list before allParams
self.pmin = np.delete(self.pmin, 1, 0)
#self.ptarget = self.orig_params
params = []
for i in range(len(self.pmin)):
params.append(
bpop.parameters.Parameter(data[i][0],
bounds=(self.pmin[i], self.pmax[i])))
self.params = params
self.weights = opt_weight_list
self.opt_stim_list = [e.decode('ascii') for e in opt_stim_name_list]
self.target_volts_list = np.array(
[target_volts_hdf5[s][:] for s in self.opt_stim_list])
self.objectives = [
bpop.objectives.Objective('Weighted score functions')
]
# AP tune stuff not currently being used
self.ap_tune_stim_name = ap_tune_stim_name
self.ap_tune_weight = ap_tune_weight
self.ap_tune_target = target_volts_hdf5[self.ap_tune_stim_name][:]
self.dts = []
def __init__(self):
"""Constructor"""
params_ = nrnUtils.readParamsCSV(paramsCSV)
super(hoc_evaluator, self).__init__()
self.orig_params = orig_params
# code below more specific to neuroGPU
data = np.genfromtxt(params_table,delimiter=',',names=True)
self.pmin = data[0]
print(self.pmin, "PMIN")
self.pmax = data[1]
self.ptarget = data[2]
params = []
for i in range(len(self.pmin)):
params.append(bpop.parameters.Parameter('p' + str(i), bounds=(self.pmin[i],self.pmax[i])))
self.params = params
#print("Params to optimize:", [(name, minval, maxval) for name, minval, maxval in params_])
self.weights = opt_weight_list
self.opt_stim_list = [e.decode('ascii') for e in opt_stim_name_list]
#self.objectives = [bpop.objectives.Objective('Weighted score functions')]
self.objectives = np.repeat(bpop.objectives.Objective('Weighted score \
functions'),4) #for 16 stims
def __init__(self):
"""Constructor"""
data = nrnUtils.readParamsCSV(paramsCSV)
super(hoc_evaluator, self).__init__()
self.orig_params = orig_params
self.opt_ind = np.array(params_opt_ind)
data = np.array([data[i] for i in self.opt_ind])
self.orig_params = orig_params
self.pmin = np.array((data[:,2]), dtype=np.float64)
self.pmax = np.array((data[:,3]), dtype=np.float64)
# make this a function
self.fixed = {}
self.params = []
for param_idx in range(len(self.orig_params)):
if param_idx in self.opt_ind:
idx = np.where(self.opt_ind == param_idx)
if np.isclose(self.orig_params[param_idx],self.pmin[idx],rtol=.000001) and np.isclose(self.pmin[idx],self.pmax[idx],rtol=.000001):
#self.fixed[param_idx] = self.orig_params[param_idx]
self.params.append(bpop.parameters.Parameter(self.orig_params[param_idx], bounds=(self.pmin[idx][0]*.999999,self.pmax[idx][0]*1.00001)))
print(" opt but fixed idx : ", (self.orig_params[param_idx], self.pmin[idx][0]*.999999,self.pmax[idx][0]*1.00001))
else:
print("USING: ", self.opt_ind[idx[0]],self.orig_params[param_idx], (self.pmin[idx],self.pmax[idx]))
counter +=1
self.params.append(bpop.parameters.Parameter(self.orig_params[param_idx], bounds=(self.pmin[idx][0],self.pmax[idx][0]))) # this indexing is annoying... pmax and pmin weird shape because they are numpy arrays, see idx assignment on line 125... how can this be more clear
else:
print("FIXED: ", self.orig_params[param_idx], " idx : ", param_idx)
self.fixed[param_idx] = self.orig_params[param_idx]
self.params.append(bpop.parameters.Parameter(self.orig_params[param_idx], bounds=(self.pmin[idx][0],self.pmax[idx][0]))) # this indexing is annoying... pmax and pmin weird shape because they are numpy arrays, see idx assignment on line 125... how can this be more clear
else:
print(self.orig_params[param_idx], " idx : ", param_idx)
self.fixed[param_idx] = self.orig_params[param_idx]
def __init__(self):
"""Constructor"""
data = nrnUtils.readParamsCSV(paramsCSV)
super(hoc_evaluator, self).__init__()
self.orig_params = orig_params
self.opt_ind = params_opt_ind
data = np.array([data[i] for i in self.opt_ind])
#print(orig_params, "orig params")
self.orig_params = orig_params
#print(1/0)
self.pmin = np.array((data[:,1]), dtype=np.float64)
# THIS IS A SHORT TERM FIX, need to fix it in evaluate with lists
self.pmax = np.array((data[:,2]), dtype=np.float64)
#testing
#self.pmin = self.orig_params - .000000001
#self.pmax = self.orig_params + .000001
self.pmax = np.delete(self.pmax, 1, 0)
self.pmin = np.delete(self.pmin, 1, 0)
#old
#self.pmax[1] = self.pmax[1] + .000001 # need to remove it from optimization and add it back in for mapping
#self.pmin[1] = self.pmin[1] - .000001 #TODO, there are some issues with this param
self.ptarget = self.orig_params
params = []
for i in range(len(self.pmin)):
# params.append(bpop.parameters.Parameter(data[i][0], bounds=(np.abs(self.ptarget[i])-.000000000001,np.abs(self.ptarget[i])+.0000000000001)))
#testing
params.append(bpop.parameters.Parameter(data[i][0], bounds=(self.pmin[i],self.pmax[i])))
self.params = params
#to reset mapping
#old
# data = np.genfromtxt(params_table,delimiter=',',names=True)
# self.pmin = data[0]
# print(self.pmin, "PMIN")
# self.pmax = data[1]
# self.ptarget = data[2]
# params = []
# for i in range(len(self.pmin)):
# params.append(bpop.parameters.Parameter('p' + str(i), bounds=(self.pmin[i],self.pmax[i])))
# self.params = params
self.weights = opt_weight_list
self.opt_stim_list = [e.decode('ascii') for e in opt_stim_name_list]
print("Init target volts")
self.target_volts_list = np.abs([target_volts_hdf5[s][:] for s in self.opt_stim_list])
#print("Params to optimize:", params[1])
self.objectives = [bpop.objectives.Objective('Weighted score functions')]
self.ap_tune_stim_name = ap_tune_stim_name
self.ap_tune_weight = ap_tune_weight
self.ap_tune_target = target_volts_hdf5[self.ap_tune_stim_name][:]
self.dts = []
def __init__(self):
"""Constructor"""
params_ = nrnUtils.readParamsCSV(paramsCSV)
super(hoc_evaluator, self).__init__()
self.opt_ind = params_opt_ind
params_ = [params_[i] for i in self.opt_ind]
self.orig_params = orig_params
self.params = [
bpop.parameters.Parameter(name, bounds=(minval, maxval))
for name, minval, maxval in params_
]
print("Params to optimize:",
[(name, minval, maxval) for name, minval, maxval in params_])
print("Orig params:", self.orig_params)
self.weights = opt_weight_list
self.opt_stim_list = [e.decode('ascii') for e in opt_stim_name_list]
self.objectives = [
bpop.objectives.Objective('Weighted score functions')
]
print("Init target volts")
self.target_volts_list = run_model(orig_params, self.opt_stim_list)
def __init__(self):
"""Constructor"""
data = nrnUtils.readParamsCSV(paramsCSV)
super(hoc_evaluator, self).__init__()
self.orig_params = orig_params
self.opt_ind = np.array(params_opt_ind)
data = np.array([data[i] for i in self.opt_ind])
realData = nrnUtils.readParamsCSV(
"../../params/params_bbp_peeling_description.csv")
realOrig = np.array((np.array(realData)[:, 1]), dtype=np.float64)
self.orig_params = orig_params
self.pmin = np.array((data[:, 2]), dtype=np.float64)
self.pmax = np.array((data[:, 3]), dtype=np.float64)
# make this a function
self.fixed = {}
self.params = []
counter = 0
for param_idx in range(len(self.orig_params)):
if param_idx in self.opt_ind:
idx = np.where(self.opt_ind == param_idx)
if np.isclose(
self.orig_params[param_idx], self.pmin[idx],
rtol=.000001) and np.isclose(
self.pmin[idx], self.pmax[idx], rtol=.000001):
#self.fixed[param_idx] = self.orig_params[param_idx]
self.params.append(
bpop.parameters.Parameter(
self.orig_params[param_idx],
bounds=(self.pmin[idx][0] * .999999,
self.pmax[idx][0] * 1.00001)))
print(" opt but fixed idx : ",
(self.orig_params[param_idx], self.pmin[idx][0] *
.999999, self.pmax[idx][0] * 1.00001))
counter += 1
else:
print("USING: ", self.opt_ind[idx[0]],
self.orig_params[param_idx],
(self.pmin[idx], self.pmax[idx]))
print(param_idx, idx)
counter += 1
self.params.append(
bpop.parameters.Parameter(self.orig_params[param_idx],
bounds=(self.pmin[idx][0],
self.pmax[idx][0]))
) # this indexing is annoying... pmax and pmin weird shape because they are numpy arrays, see idx assignment on line 125... how can this be more clear
else:
print("FIXED: ", self.orig_params[param_idx], " idx : ",
param_idx)
self.fixed[param_idx] = self.orig_params[param_idx]
print("NUM FREE PARAMS :", counter)
self.weights = opt_weight_list
self.opt_stim_list = [e.decode('ascii') for e in opt_stim_name_list]
self.objectives = [
bpop.objectives.Objective('Weighted score functions')
]
self.target_volts_list = self.make_target_volts(
realOrig, self.opt_stim_list)
self.dts = []
scores_path = '../scores/'
objectives_file = h5py.File('../objectives/multi_stim_bbp_full.hdf5', 'r')
opt_weight_list = objectives_file['opt_weight_list'][:]
opt_stim_name_list = objectives_file['opt_stim_name_list'][:]
score_function_ordered_list = objectives_file['ordered_score_function_list'][:]
stims_path = '../stims/stims_full.hdf5'
stim_file = h5py.File(stims_path, 'r')
#target_volts_path = './target_volts/allen_data_target_volts_10000.hdf5'
#target_volts_hdf5 = h5py.File(target_volts_path, 'r')
#params_opt_ind = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13]
params_opt_ind = np.arange(24)
model_dir = '..'
data_dir = model_dir + '/Data/'
run_dir = '../bin'
vs_fn = '/tmp/Data/VHotP'
data = nrnUtils.readParamsCSV(paramsCSV)
#self.opt_ind = params_opt_ind
data = np.array([data[i] for i in params_opt_ind])
pmin = np.array((data[:, 2]), dtype=np.float64)
pmax = np.array((data[:, 3]), dtype=np.float64)
# make this a function
fixed = {}
params = []
for param_idx in range(len(orig_params)):
if np.isclose(orig_params[param_idx], pmin[param_idx],
rtol=.001) and np.isclose(
pmin[param_idx], pmax[param_idx], rtol=.001):
print(orig_params[param_idx], " idx : ", param_idx)
fixed[param_idx] = orig_params[param_idx]
else:
pass