TL Review interfaces.ipynb

[tiekneeshrimp]

1. Spelling error

• Issue : spell checking for "varibales" into "variables
• Place: Under "Calibrate interface" heading documentation

2. Graph sections disjointed

• Issue: Cases, Death, Hospitalizations graph areas are not overlapping
• Place: Under code chunk [11] , "Use calibrated parameter estimates in ensemble_sample to sample from calibrated ensemble model
• Code:

train_end_point = 3.3   # Use train_end_point = None if there is no calibration
if train_end_point is None:
    end_time_ensemble = 28.0
else:
    end_time_ensemble = train_end_point + 28.0
calibrated_ensemble_result = pyciemss.ensemble_sample(model_paths, solution_mappings, end_time_ensemble, logging_step_size_ensemble, num_samples, 
                                                      start_time=start_time, inferred_parameters=parameter_estimates)
display(calibrated_ensemble_result['data'].head())

# Plot the ensemble result for cases, hospitalizations, and deaths
nice_labels={"dead_state": "Deaths", 
                 "hospitalized_state": "Hospitalizations",
                 "infected_state": "Cases"
                }
schema = plots.trajectories(calibrated_ensemble_result["data"], 
                           keep=["infected_state", "hospitalized_state", "dead_state"], 
                           relabel=nice_labels,
                          )
plots.save_schema(schema, "_schema.json")
plots.ipy_display(schema, dpi=150)

-Expectation: Sections of 'cases', 'deaths', and 'hostpitalizations' should maybe overlap or touch instead of being individual chunks?
-Actual Results:

3. Optimizing interface for optimal start time failed -- but still generated a graph?

• Issue: Seems that the optimization did not succeed? "success = False". Still generated a graph, but the solution not satisfying constraints seemed to come up as a warning for all following codes that used the optimized start time
• Place: Under code chunk [19] , "Optimize interface for optimizing start time"

initial_guess_interventions = 0.0
bounds_interventions = [[start_time], [end_time]]

risk_bound = 300.0
qoi = lambda x: obs_nday_average_qoi(x, observed_params, 1)
objfun = lambda x: -x

static_parameter_interventions = start_time_objective(
    param_name = intervened_params,
    param_value = torch.tensor([0.15]),
)
opt_result = pyciemss.optimize(
    model3,
    end_time,
    logging_step_size,
    qoi,
    risk_bound,
    static_parameter_interventions,
    objfun,
    initial_guess_interventions=initial_guess_interventions,
    bounds_interventions=bounds_interventions,
    start_time=0.0,
    n_samples_ouu=int(1e2),
    maxiter=1,
    maxfeval=10,
    solver_method="euler",
    solver_options={"step_size": logging_step_size/2},
)

print(f'Optimal policy:', opt_result["policy"])
print(opt_result)

-Expectation: i think the success message should be "True", and lowest_optimization_result: message: would have a solution output
-Actual Results: