def get_sim(type, output, dh=None):
model = system.get_model()
if type == 'h**o':
model.collapse(['ii', 'ip'])
name = '$G = 1$ (Homogeneous)'
dt = 0.2
if type == 'zeta':
dhmax = 1.0 / model.params['mu']
model.params['dur'].update(
[dh, min(dh * 5, dhmax),
min(dh * 30, dhmax)])
# z1 = (1/model.params['dur'].iselect(ii='H',ki='M') - model.params['mu'])/2
# z2 = (1/model.params['dur'].iselect(ii='M',ki='M') - model.params['mu'])/2
# z3 = (1/model.params['dur'].iselect(ii='L',ki='M') - model.params['mu'])/2
# model.params['zeta'].update(z1,ii='H',ip='M')
# model.params['zeta'].update(z1,ii='H',ip='L')
# model.params['zeta'].update(z2,ii='M',ip='L')
# model.params['zeta'].update(z2,ii='M',ip='H')
# model.params['zeta'].update(z3,ii='L',ip='M')
# model.params['zeta'].update(z3,ii='L',ip='H')
model.params['zeta'].update(np.nan)
# model.params['pe'].update(np.nan)
name = '$G = 3, \\delta_H = {:.03f}$'.format(model.params['dur'][0, 0])
dt = min(dh * 0.8, 0.2)
t = system.get_t(dt=dt, tmin=0, tmax=200)
sim = system.get_simulation(model, t=t)
sim.init_outputs(
system.get_outputs(
spaces=sim.model.spaces,
select=sim.model.select,
t=sim.t,
names=[output],
))
return sim, name
def run_fit():
t = system.get_t(tmax=500)
sims = odict([
('Turnover', get_sim('full', t=t)),
('No Turnover', get_sim('no-turnover', t=t)),
])
for name, sim in sims.items():
sim.init_outputs(
system.get_outputs(spaces=sim.model.spaces,
select=sim.model.select,
t=sim.t,
names=['prevalence']))
sim.solve()
targets = system.get_targets(
sim.model.spaces,
sim.model.select,
sim.outputs,
t=sim.t,
)
if config.model == 'mort':
sim.model.params['C'].update([25, 10,
5]) # HACK to get low prev > 0
calsim = calibration.CalibrationSim(
name,
sim=sim,
targets=targets,
verbose=True,
)
calsim.optimize(ftol=1e-3, plot='tmp.png')
if config.save:
utils.savejson(fname_fit(name), calsim.fitted_params().todict())
else:
print(dict(calsim.fitted_params().todict()))
def run_sim(sim, outputs=None):
outputs = outputs if outputs is not None else []
sim.init_outputs(
system.get_outputs(spaces=sim.model.spaces,
select=sim.model.select,
t=sim.t,
names=outputs))
return sim.solve()
def make_figs():
phis = list(sensitivity.iter_phi())
for label,phi in [
('low', phis[n]),
('med', phis[int((config.N-1)/2)]),
('high', phis[config.N-n-1]),
('extreme',10),
]:
specs = system.get_specs()
model = system.get_model()
sim = sensitivity.get_sim(phi,0.1)
sim.init_outputs(system.get_outputs(
spaces = sim.model.spaces,
select = sim.model.select,
t = sim.t,
names = [out]
))
if label == 'extreme':
sim.update_params(dict(ibeta=0.038))
sim.solve()
for name in ['high','low']:
make_pie(sim,name,phi)
make_tikz(label,phi)
def get_sim(phi, tau, tmax=200):
model = system.get_model()
dmax = 1.0 / model.params['mu']
dh = 1 / phi
model.params['dur'].update(dh, ii=['H'])
model.params['dur'].update(dh + 0.30 * (dmax - dh), ii=['M'])
model.params['dur'].update(np.nan, ii=['L'])
p1 = (1 / model.params['dur'].iselect(ii=['H'], ki=['M']) -
model.params['mu']) / 2
model.params['phi'].update(p1, ii=['H'], ip=['M'])
model.params['phi'].update(p1, ii=['H'], ip=['L'])
model.params['tau'].update(tau)
dt = min(dh * 0.8, 0.5)
t = system.get_t(dt=dt, tmin=0, tmax=tmax)
sim = system.get_simulation(model, t=t)
sim.init_outputs(
system.get_outputs(
spaces=sim.model.spaces,
select=sim.model.select,
t=sim.t,
names=OUTPUTS + ['C', 'X', 'tip'],
))
return sim
# define true homogeneous model
models['h**o'].collapse(['ii'])
# define high zeta model
models['zeta'].params['pe'].update(np.nan)
models['zeta'].params['dur'].update([0.01, 0.05, 0.8])
models['zeta'].params['zeta'].update(np.nan, ii='H', ip='L')
sims = odict([
('Homogeneous', system.get_simulation(models['h**o'], dt=dt)),
('High Turnover', system.get_simulation(models['zeta'], dt=dt)),
])
for (name, sim), clr in zip(sims.items(), [[1, 0, 0], [0, 0, 1]]):
sim.init_outputs(
system.get_outputs(
spaces=sim.model.spaces,
select=sim.model.select,
t=system.get_t(dt=dt),
names=[output],
))
for param in ['pe', 'dur', 'zeta']:
print('{} = \n{}'.format(param, np.around(sim.model.params[param], 6)),
flush=True)
sim.solve(eqfun=lambda s, t: print(t, flush=True))
y = modelutils.taccum(sim.outputs[output].iselect(sim.model.select['all']))
plt.plot(sim.t, y, color=clr)
sim = None
plt.ylim([0, 0.1])
plt.ylabel('Prevalence (overall)')
plt.xlabel('t (years)')
plt.legend(sims.keys())
plt.savefig('infinite-zeta-tmp.eps')