for i in range(n_models):
# create a model randomly from all channels
weights_model = problem.generator(prng, None)
for j, dt in enumerate(dts):
print 'run '+'model: ' + str(i) + ' dt: ' +str(dt)
# change dt
data_newdt = change_dt(dt, problem.data)
problem.simulation_params = extract_simulation_params(data_newdt)
# run simulation
problem.update_cell(weights_model)
currents = [problem.cell.soma.record_from(channel_list[k], 'i'+ion_list[k], pos=.5) for k in range(len(channel_list))]
v_newdt, t_newdt = run_simulation(problem.cell, **problem.simulation_params)
#pl.figure()
#pl.plot(t_newdt, v_newdt)
#pl.show()
# compute parameter
dvdt_newdt = np.concatenate((np.array([(v_newdt[1]-v_newdt[0])/dt]), np.diff(v_newdt)/dt))
i_newdt = data_newdt.i.values
celsius = problem.simulation_params['celsius']
# get currents
candidate = np.ones(len(problem.path_variables)) # gbars should be 1
problem.update_cell(candidate)
currents_newdt = currents_given_v(v_newdt, t_newdt, problem.cell.soma, channel_list, ion_list, celsius)
for i in range(len(merge_points)-1):
y_plot = y[merge_points[i]:merge_points[i+1]]
X_plot = X[merge_points[i]:merge_points[i+1], :]
t_plot = t_exp[merge_points[i]:merge_points[i+1]]
plot_fit(y_plot, X_plot, weights, t_plot, channel_list, save_dir=save_dir+str(i))
# save
np.savetxt(save_dir+'/best_candidate_'+str(trial)+'.txt', weights)
np.savetxt(save_dir+'/error_'+str(trial)+'.txt', np.array([residual]))
# simulate
#cm = weights[-1]
for i, w in enumerate(weights[:]):
keys = ['soma', '0.5', channel_list[i], 'gbar']
if channel_list[i] == 'pas':
keys = ['soma', '0.5', channel_list[i], 'g']
problem.cell.update_attr(keys, w)
elif 'ion' in channel_list[i]:
keys = None
else:
keys = ['soma', '0.5', channel_list[i], 'gbar']
problem.cell.update_attr(keys, w)
from optimization.simulate import run_simulation
v, t = run_simulation(problem.cell, **problem.simulation_params)
import matplotlib.pyplot as pl
pl.figure()
pl.plot(t, problem.data.v, 'k')
pl.plot(t, v, 'r')
pl.show()