def get_gridspec(fig: plt.Figure, num: int, return_list=True):
if num > 12:
logger.warning(
'[add_gridspec]: got request for more than 12 plots, only returning 12'
)
num = 12
if num != 0:
shape_dict = {
1: (1, 1),
2: (2, 1),
3: (2, 2),
5: (3, 2),
7: (3, 3),
10: (4, 3)
}
arr = np.array(list(shape_dict.keys())) # array of index
key = arr[arr <= num].max(
) # largest element lower than num (i.e. previous size which fits required num
shape = shape_dict[key]
gs = fig.add_gridspec(*shape)
if return_list is True:
gs = [gs[i, j] for i in range(shape[0]) for j in range(shape[1])]
else:
if return_list is True:
gs = []
else:
gs = None
return gs
def __abc_smc_plotting(fig: plt.Figure, y_obs: [[float]],
priors: ["stats.Distribution"],
fitting_model: Models.Model, model_hat: Models.Model,
accepted_params: [[float]],
weights: [float]) -> plt.Figure:
n_params = (fitting_model.n_params - 2) if (
type(fitting_model) is Models.SIRModel) else fitting_model.n_params
n_rows = max([1, np.lcm(n_params, fitting_model.dim_obs)])
gs = fig.add_gridspec(n_rows, 2)
# plot fitted model
row_step = n_rows // fitting_model.dim_obs
for i in range(fitting_model.dim_obs):
ax = fig.add_subplot(gs[i * row_step:(i + 1) * row_step, -1])
y_obs_dim = [y[i] for y in y_obs]
Plotting.plot_accepted_observations(ax,
fitting_model.x_obs,
y_obs_dim, [],
model_hat,
dim=i)
row_step = n_rows // n_params
if (type(fitting_model) is Models.SIRModel):
for i in range(2, fitting_model.n_params):
ax = fig.add_subplot(gs[(i - 2) * row_step:(i - 2 + 1) * row_step,
0])
name = "theta_{}".format(i)
accepted_parameter_values = [theta[i] for theta in accepted_params]
Plotting.plot_parameter_posterior(
ax,
name,
accepted_parameter_values,
predicted_val=model_hat.params[i],
prior=priors[i],
dim=i,
weights=weights)
else:
for i in range(fitting_model.n_params):
ax = fig.add_subplot(gs[i * row_step:(i + 1) * row_step, 0])
name = "theta_{}".format(i)
parameter_values = [theta[i] for theta in accepted_params]
Plotting.plot_parameter_posterior(
ax,
name,
accepted_parameter_values,
predicted_val=model_hat.params[i],
prior=priors[i],
dim=i,
weights=weights)
return fig
def __abc_mcmc_plotting(fig: plt.Figure, y_obs: [[float]],
priors: ["stats.Distribution"],
fitting_model: Models.Model, model_hat: Models.Model,
accepted_summary_vals: [[float]],
accepted_params: [[float]]) -> plt.Figure:
n_simple_ss = sum(
len(s) == 1 for s in accepted_summary_vals[0]
) # number of summary stats which map to a single dimension
n_cols = 3 if (n_simple_ss == 0) else 4
n_params = (fitting_model.n_params - 2) if (
type(fitting_model) is Models.SIRModel) else fitting_model.n_params
n_rows = max([
1,
np.lcm.reduce([n_params,
max(1, n_simple_ss), fitting_model.dim_obs])
])
gs = fig.add_gridspec(n_rows, n_cols)
# plot fitted model
row_step = n_rows // fitting_model.dim_obs
for i in range(fitting_model.dim_obs):
ax = fig.add_subplot(gs[i * row_step:(i + 1) * row_step, -1])
y_obs_dim = [y[i] for y in y_obs]
Plotting.plot_accepted_observations(ax,
fitting_model.x_obs,
y_obs_dim, [],
model_hat,
dim=i)
# plot traces & posteriors
row_step = n_rows // n_params
if (type(fitting_model) is Models.SIRModel):
for i in range(2, fitting_model.n_params):
name = "Theta_{}".format(i)
accepted_vals = [x[i] for x in accepted_params]
# Chain trace
ax = fig.add_subplot(gs[(i - 2) * row_step:((i - 2) + 1) *
row_step, 0])
Plotting.plot_MCMC_trace(ax,
name,
accepted_parameter=accepted_vals,
predicted_val=model_hat.params[i])
# Posteriors
ax = fig.add_subplot(gs[(i - 2) * row_step:((i - 2) + 1) *
row_step, 1])
Plotting.plot_parameter_posterior(
ax,
name,
accepted_parameter=accepted_vals,
predicted_val=model_hat.params[i],
prior=priors[i],
dim=i)
else:
for i in range(fitting_model.n_params):
name = "Theta_{}".format(i)
accepted_vals = [x[i] for x in THETAS]
# Chain trace
ax = fig.add_subplot(gs[i * row_step:(i + 1) * row_step, 0])
Plotting.plot_MCMC_trace(ax,
name,
accepted_parameter=accepted_vals,
predicted_val=model_hat.params[i])
# Posteriors
ax = fig.add_subplot(gs[i * row_step:(i + 1) * row_step, 1])
Plotting.plot_parameter_posterior(
ax,
name,
accepted_parameter=accepted_vals,
predicted_val=model_hat.params[i],
prior=priors[i],
dim=i)
# plot summary vals
row = 0
row_step = n_rows // n_simple_ss
for i in range(len(accepted_summary_vals[0])):
if (len(accepted_summary_vals[0][i]) == 1):
name = "s_{}".format(i)
ax = fig.add_subplot(gs[row * row_step:(row + 1) * row_step, 2])
row += 1
accepted_vals = [s[i][0] for s in accepted_summary_vals]
Plotting.plot_summary_stats(ax,
name,
accepted_s=accepted_vals,
s_obs=s_obs[i],
s_hat=s_hat[i],
dim=i)
return fig
def __abc_rejection_plotting(fig: plt.Figure, y_obs: [[float]], priors: [
"stats.Distribution"
], fitting_model: Models.Model, model_hat: Models.Model,
accepted_summary_vals: [[float]],
accepted_observations: [[float]],
accepted_params: [[float]]) -> plt.Figure:
# plot results
n_simple_ss = sum(
len(s) == 1 for s in accepted_summary_vals[0]
) # number of summary stats which map to a single dimension
n_cols = 2 if (n_simple_ss == 0) else 3
n_params = (fitting_model.n_params - 2) if (
type(fitting_model) is Models.SIRModel) else fitting_model.n_params
n_rows = max([
1,
np.lcm.reduce([n_params,
max(1, n_simple_ss), fitting_model.dim_obs])
])
# plot accepted obervations for each dimension
gs = fig.add_gridspec(n_rows, n_cols)
# plot accepted observations (each dimension separate)
row_step = n_rows // fitting_model.dim_obs
for i in range(fitting_model.dim_obs):
ax = fig.add_subplot(gs[i * row_step:(i + 1) * row_step, -1])
y_obs_dim = [y[i] for y in y_obs]
accepted_obs_dim = [[y[i] for y in obs]
for obs in accepted_observations]
Plotting.plot_accepted_observations(ax,
fitting_model.x_obs,
y_obs_dim,
accepted_obs_dim,
model_hat,
dim=i)
# plot posterior for each parameter
row_step = n_rows // n_params
if (type(fitting_model) is Models.SIRModel):
for i in range(2, fitting_model.n_params):
name = "Theta_{}".format(i)
ax = fig.add_subplot(gs[(i - 2) * row_step:(i - 2 + 1) * row_step,
0])
accepted_vals = [x[i] for x in accepted_params]
Plotting.plot_parameter_posterior(
ax,
name,
accepted_parameter=accepted_vals,
predicted_val=model_hat.params[i],
prior=priors[i],
dim=i)
else:
for i in range(fitting_model.n_params):
name = "Theta_{}".format(i)
ax = fig.add_subplot(gs[i * row_step:(i + 1) * row_step, 0])
accepted_vals = [x[i] for x in accepted_params]
Plotting.plot_parameter_posterior(
ax,
name,
accepted_parameter=accepted_vals,
predicted_val=model_hat.params[i],
prior=priors[i],
dim=i)
# plot histogram of each summary statistic value
row = 0
if (n_simple_ss != 0):
row_step = n_rows // n_simple_ss
for i in range(len(summary_stats)):
if (len(accepted_summary_vals[0][i]) == 1):
name = "s_{}".format(i)
ax = fig.add_subplot(gs[row * row_step:(row + 1) * row_step,
1])
row += 1
accepted_vals = [s[i][0] for s in accepted_summary_vals]
Plotting.plot_summary_stats(ax,
name,
accepted_s=accepted_vals,
s_obs=s_obs[i],
s_hat=s_hat[i],
dim=i)
return fig
