population_file = pd.read_csv(path, names=['generation', 'index', 'fitness', 'candidate'], header=None)
# estimate error landscape
with open('./results/params.json') as f:
params = json.load(f)
p1_range = np.arange(params['lower_bound'], params['upper_bound']+0.0001, 0.01)
p2_range = np.arange(params['lower_bound'], params['upper_bound']+0.0001, 0.01)
if os.path.isfile('./results/error_landscape.npy'):
error = np.load('./results/error_landscape.npy')
else:
candidates = list()
for i, p1 in enumerate(p1_range):
for j, p2 in enumerate(p2_range):
candidates.append([p1, p2])
problem = CellFitProblem(params)
error = problem.evaluator(candidates, [])
error = np.reshape(error, (len(p1_range), len(p2_range)))
with open('./results/error_landscape.npy', 'w') as f:
np.save(f, error)
fig = plt.figure()
ax = p3.Axes3D(fig)
ax.set_xlim3d([0.0, 1.0])
ax.set_xlabel('X')
ax.set_ylim3d([0.0, 1.0])
ax.set_ylabel('Y')
ax.set_zlabel('Z')
ax.view_init(elev=20, azim=50)
X, Y = np.meshgrid(p1_range, p2_range)
surf = ax.plot_surface(X, Y, error.T, rstride=1, cstride=1, cmap=cm.coolwarm, alpha=0.6,
linewidth=0, antialiased=False)
]
params = {
'maximize': False,
'normalize': False,
'model_dir': '../../model/cells/toymodel1.json',
'mechanism_dir': '../../model/channels/schmidthieber',
'variables': variables,
'data_dir': '../../data/toymodels/toymodel1/ramp.csv',
'get_var_to_fit': 'get_v',
'fitnessweights': [1.0],
'errfun': 'rms',
'insert_mechanisms': True
}
problem = CellFitProblem(**params)
dt = problem.simulation_params['dt']
time_cut = 20
problem.simulation_params['tstop'] = time_cut # TODO
problem.simulation_params['i_amp'] = problem.simulation_params['i_amp'][:time_cut/dt+1]
subsample = 5
# initialize pseudo random number generator
seed = time()
prng = Random()
prng.seed(seed)
# initialize data
t = np.arange(0, problem.simulation_params['tstop']+dt, dt)
i_inj = problem.simulation_params['i_amp']
data = np.zeros((n_data, len(t)/subsample, 2)) # inputs are the trace of v and i_inj
