def initial_guess(model):
ax = plt.subplot(111)
t_ref = model.reference_data[:, 0]
y_ref = model.reference_data[:, 1:]
plt.title('Initial model behavior and reference')
plt.plot(t_ref, y_ref, ls='--', linewidth=2)
initial_model = deepcopy(nemf.forward_model(model, t_eval=t_ref))
ax = draw_model_output(ax, initial_model)
ax.title.set_text('Initial model guess')
plt.show()
return ax
def test_forward_model_v1(model_minimal_pkl):
nemf.forward_model(model_minimal_pkl)
def test_forward_model_v6(model_npzd_stable_refed_pkl):
nemf.forward_model(model_npzd_stable_refed_pkl)
def test_forward_model_v5(model_npzd_osci_refed_pkl):
nemf.forward_model(model_npzd_osci_refed_pkl)
def test_forward_model_v4(model_npzd_stable_pkl):
nemf.forward_model(model_npzd_stable_pkl)
def test_forward_model_v3(model_npzd_osci_pkl):
nemf.forward_model(model_npzd_osci_pkl)
def forward_pickler(path,name): model = load_model(path) model = nemf.forward_model(model) write_pickle(model,name)
# NPZD Example # import the module import nemf # provide the path of model/system configuration model_path = 'example_files/NPZD/NPZD_model.yml' ref_data_path = 'example_files/NPZD/NPZD_ref_oscillating.csv' # load the model configuration model_config = nemf.model_class(model_path, ref_data_path) # visualise the model configuration to check for errors nemf.interaction_graph(model_config) # for a simple time evolution of the model call: output_dict = nemf.forward_model(model_config) # the results of the time evolution can be visualized with: nemf.output_summary(output_dict) # if the model shall be fitted as well call: model = nemf.inverse_model(model_config) # to plot the results: nemf.output_summary(model) # writes optimized model to file model.export_to_yaml(path='optimized_model.yml')
def draw_optimization_overview(model):
""" reads the data saved in the model class and depending on this data
chooses a visualization method to present the results """
fig = plt.figure()
fig.suptitle('Results of optimization run')
if model.reference_data[:, 0][0] == np.inf:
non_steady_state = False
else:
non_steady_state = True
if not np.isnan(model.log['cost'][0]):
ax1 = plt.subplot(221)
draw_cost(ax1, model.log['cost'])
ax2 = plt.subplot(222)
draw_parameters(ax2, model.log['parameters'], model)
ax3 = plt.subplot(212)
if non_steady_state:
t_ref = model.reference_data[:, 0]
y_ref = model.reference_data[:, 1:]
plt.plot(t_ref, y_ref, ls='--', linewidth=2)
time_series = nemf.forward_model(model, t_eval=t_ref)
time_series_model = time_series.log['time_series']
ax2 = draw_model_output(ax2, model)
ax2.title.set_text('optimized model')
else: # steady-state
t_max = model.configuration['max_time_evo']
t_ref = np.linspace(0, t_max, 1000)
y_ref = model.reference_data[:, 1:]
plt.hlines(y_ref, t_ref[0], t_ref[-1], ls='--')
time_series_model = nemf.forward_model(model, t_eval=t_ref)
ax3 = draw_model_output(ax3, time_series_model)
ax3.title.set_text('optimized model')
else:
ax1 = plt.subplot(211)
draw_parameters(ax1, model.log['parameters'], model)
ax2 = plt.subplot(212)
if non_steady_state:
t_ref = model.reference_data[:, 0]
y_ref = model.reference_data[:, 1:]
colors = sns.color_palette('husl', len(model.compartment))
ax2.title.set_text('optimized model')
# plotting model output
time_series = nemf.forward_model(model, t_eval=t_ref)
time_series_model = time_series.log['time_series']
ax2 = draw_model_output(ax2, model, colors=colors)
# plotting reference data
ax2 = draw_ref_data(ax2, model, colors)
else: # steady-state
t_max = model.configuration['max_time_evo']
dt = model.configuration['dt_time_evo']
t_ref = np.arange(0, t_max, dt)
y_ref = model.reference_data[:, 1:]
plt.hlines(y_ref, t_ref[0], t_ref[-1], ls='--')
time_series_model = nemf.forward_model(model, t_eval=t_ref)
draw_model_output(ax2, time_series_model)
ax2.title.set_text('optimized model')
return fig
