def ex3_plots(instance, destination, prefix, save, animate):
plts = ukf_plots(instance, destination, prefix, save, animate)
truths = truth_parser(instance)
nan_array = nan_array_parser(instance, truths, instance.base_model)
#obs, obs_key = obs_parser(instance, True)
obs_key = obs_key_parser(instance, True)
preds = preds_parser(instance, True)
#forecasts = forecasts_parser(instance, True)
gates = np.array(instance.estimated_gates).astype('float64')
ukf_params = instance.ukf_params
#forecasts = np.vstack(instance.forecasts)
"remove agents not in model to avoid wierd plots"
truths *= nan_array
preds *= nan_array
#forecasts*= nan_array
#gates *= nan_array[::instance.sample_rate, 0::2]
"indices for unobserved agents"
subset = np.arange(truths.shape[1] // 2)
subset2 = np.repeat((2 * subset), 2)
subset2[1::2] += 1
#turn on for subset plots
#if instance.model_params["random_seed"] == 8:
# subset = np.array([1, 5, 3])
# subset2 = np.repeat((2*subset), 2)
# subset2[1::2] +=1
#plts.path_plots(obs, "Observed")
plts.pair_frame(truths, preds, obs_key, 100, destination)
error_plot(instance.true_gate, instance.estimated_gates,
instance.sample_rate)
plts.error_hist(truths[::instance.sample_rate, :],
preds[::instance.sample_rate, :], "Observed Errors")
plts.path_plots(preds[::instance.sample_rate, subset2], "Predicted")
plts.path_plots(truths[:, subset2], "True")
#plts.path_plots(forecasts[::instance.sample_rate, subset2], "Forecasts")
plts.dual_path_plots(truths[::instance.sample_rate, subset2],
preds[::instance.sample_rate,
subset2], "True and Predicted")
plts.gate_choices(gates[:, ], instance.sample_rate)
#split_gate_choices(gates[:, subset], instance.sample_rate)
if animate:
#plts.trajectories(truths, "plots/")
plts.pair_frames(truths[::instance.sample_rate],
preds[::instance.sample_rate],
obs_key[::instance.sample_rate],
truths[::instance.sample_rate].shape[0],
"../../plots/")
def ex1_plots(instance, destination, prefix, save, animate):
"""do plots for experiment 1
- extract truths, obs, ukf predictions (preds), and forecasts
- remove inactive agent measurements to prevent bias
- plot one pairs plot frame linking ukf predictions to true values by tethers
- plot population of agent median L2 errors in a histogram
- plot agent paths for observations, ukf preditions, and true values
- if animated
- plot true paths for agents in trajectories
- plot a pair plot frame for every model step and animate them together.
Parameters
------
instance : class
uks_ss class `instance` after completed run.
plot_dir, prefix : str
`plot_dir` where to save the plot to e.g. "" for this directory or
"ukf_results/ for ukf_results"
`prefix` puts some prefix in front of various picture file names.
e.g. "ukf_" or "agg_ukf_" so you can differentiate them and not overwrite
them
save , animate : bool
`save` plots or `animate` whole model run?
"""
marker_attributes = {
"markers": {
-1: "o",
0: "X",
2: "s"
},
"colours": {
-1: "black",
0: "orangered",
2: "skyblue"
},
"labels": {
-1: "Pseudo-True Positions",
0: "Unobserved Agents",
2: "Observed Agents"
}
}
plts = ukf_plots(instance, destination, prefix, save, animate,
marker_attributes)
obs, preds, truths, nan_array = instance.data_parser()
obs_key = instance.obs_key_parser()
ukf_params = instance.ukf_params
index2 = ukf_params["index2"]
forecasts = np.vstack(instance.forecasts)
obs *= nan_array[::instance.sample_rate, index2]
truths *= nan_array
preds *= nan_array
forecasts *= nan_array
"indices for unobserved agents"
not_index2 = np.array(
[i for i in np.arange(truths.shape[1]) if i not in index2])
plts.pair_frame(truths, preds, obs_key, 50, "plots/")
plts.error_hist(truths[:, index2], preds[:, index2], "Observed Errors")
plts.error_hist(truths[:, not_index2], preds[:, not_index2],
"Unobserved Errors")
plts.path_plots(obs, "Observed")
"remove nan rows to stop plot clipping"
plts.path_plots(preds[::instance.sample_rate], "Predicted")
plts.path_plots(truths, "True")
if animate:
#plts.trajectories(truths, "plots/")
plts.pair_frames(truths, forecasts, obs_key, truths.shape[0], "plots/")
def ex2_plots(instance, destination, prefix, save, animate):
"""plots for experiments 2
Parameters
------
instance : class
ukf_ss `instance` containing a finished stationsim run to publish
destination, prefix : str
`destination to save
save, animate : bool
"""
marker_attributes = {
"markers": {
-1: "o",
0: "X",
1: "^"
},
"colours": {
-1: "black",
0: "orangered",
1: "yellow"
},
"labels": {
-1: "Pseudo-Truths",
0: "Unobserved",
1: "Aggregated"
}
}
plts = ukf_plots(instance, destination, prefix, save, animate,
marker_attributes)
"pull data and put finished agents to nan"
obs, preds, truths, nan_array = instance.data_parser()
obs_key = instance.obs_key_parser()
forecasts = np.vstack(instance.forecasts)
truths *= nan_array
preds *= nan_array
forecasts *= nan_array
plts.pair_frame(truths, preds, obs_key, 50, "plots/")
plts.heatmap_frame(truths, 50, "plots/")
plts.error_hist(truths, preds, "Aggregate")
"remove nan rows to stop plot clipping"
plts.path_plots(preds[::instance.sample_rate],
"Predicted",
polygons=instance.ukf_params["poly_list"])
plts.path_plots(truths, "True")
if animate:
#plts.trajectories(truths, "plots/")
#plts.heatmap(truths,truths.shape[0], "plots/")
plts.pair_frames(truths, forecasts, np.vstack(obs_key),
truths.shape[0], "plots/")
def ex1_plots(instance, destination, prefix, save, animate):
"""do plots for experiment 1
- extract truths, obs, ukf predictions (preds), and forecasts
- remove inactive agent measurements to prevent bias
- plot one pairs plot frame linking ukf predictions to true values by tethers
- plot population of agent median L2 errors in a histogram
- plot agent paths for observations, ukf preditions, and true values
- if animated
- plot true paths for agents in trajectories
- plot a pair plot frame for every model step and animate them together.
Parameters
------
instance : class
uks_ss class `instance` after completed run.
plot_dir, prefix : str
`plot_dir` where to save the plot to e.g. "" for this directory or
"ukf_results/ for ukf_results"
`prefix` puts some prefix in front of various picture file names.
e.g. "ukf_" or "agg_ukf_" so you can differentiate them and not overwrite
them
save , animate : bool
`save` plots or `animate` whole model run?
"""
plts = ukf_plots(instance, destination, prefix, save, animate)
truths = truth_parser(instance)
nan_array= nan_array_parser(instance, truths, instance.base_model)
#obs, obs_key = obs_parser(instance, True)
obs_key = obs_key_parser(instance, True)
preds = preds_parser(instance, True)
#forecasts = forecasts_parser(instance, True)
ukf_params = instance.ukf_params
index2 = ukf_params["index2"]
"remove agents not in model to avoid wierd plots"
#obs *= nan_array
truths *= nan_array
preds *= nan_array
#forecasts*= nan_array
"indices for unobserved agents"
not_index2 = np.array([i for i in np.arange(truths.shape[1]) if i not in index2])
plts.pair_frame(truths, preds, obs_key, 10, destination)
plts.error_hist(truths[::instance.sample_rate,index2],
preds[::instance.sample_rate,index2],"Observed Errors")
if len(not_index2)>0:
plts.error_hist(truths[::instance.sample_rate, not_index2],
preds[::instance.sample_rate, not_index2],"Unobserved Errors")
#plts.path_plots(obs[::instance.sample_rate] , "Observed")
plts.path_plots(preds[::instance.sample_rate], "Predicted")
plts.path_plots(truths, "True")
#plts.path_plots(forecasts[::instance.sample_rate], "Forecasts")
if animate:
#plts.trajectories(truths, "plots/")
plts.pair_frames(truths, preds, obs_key,
truths.shape[0], "../../plots/")