# n_step = int(5e5)
# hyper_prob0 = np.array([0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.45, 0.6, 0.8])
# hyper_prob0 = np.array([0.5, 0.5, 0.55, 0.5, 0.55, 0.3, 0.5, 0.4, 0.3, 0.45, 0.6, 0.8])
hyper_prob0 = np.loadtxt(hyperic)
hyper_stepsize = (2.0 * 1e-4) * np.ones(3 * n_pop)
local_prob0 = 0.5 * np.ones(n_fixm + 2 * n_varm)
local_stepsize = (2.0 * 1e-4) * np.ones(n_fixm + 2 * n_varm)
hyper_prob_chain, hyper_chain, local_prob_chain, local_chain, loglike_chain, repeat_chain, stop_step = hbm_joint_cdf(
hyper_prob0,
hyper_stepsize,
local_prob0,
local_stepsize,
n_step,
inverse_hyper,
inverse_local,
loglike_func,
data=[dat_fixm, dat_varm],
trial_upbound=100 * n_step,
)
if np.any(loglike_chain == -np.inf):
stop_step = np.where(loglike_chain == -np.inf)[0][0]
else:
stop_step = n_step
if repeat[stop_step - 1] == 0:
repeat[stop_step - 1] = 1
### save
#n_step = int(5e5) #hyper_prob0 = np.array([0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.45, 0.6, 0.8]) #hyper_prob0 = np.array([0.5, 0.5, 0.55, 0.5, 0.55, 0.3, 0.5, 0.4, 0.3, 0.45, 0.6, 0.8]) hyper_prob0 = np.loadtxt(input+'hyper_prob_top.out')[0-top,:] hyper_stepsize = (2.*1e-4) * np.ones(3*n_pop) local_prob0 = np.loadtxt(input+'local_prob_top.out')[0-top,:] local_stepsize = (2.*1e-4) * np.ones(n_fixm + 2*n_varm) np.savetxt(dir+'hyper_prob_start.out', hyper_prob0) np.savetxt(dir+'local_prob_start.out', local_prob0) hyper_prob_chain, hyper_chain, local_prob_chain, local_chain, \ loglike_chain, repeat_chain, stop_step = \ hbm_joint_cdf(hyper_prob0, hyper_stepsize, local_prob0, local_stepsize, n_step,\ inverse_hyper, inverse_local, \ loglike_func, data = [dat_fixm, dat_varm], \ trial_upbound = 20*n_step) ### save np.savetxt(dir+'hyper_prob.out',hyper_prob_chain[:stop_step,:]) np.savetxt(dir+'hyper.out',hyper_chain[:stop_step,:]) np.savetxt(dir+'loglike.out',loglike_chain[:stop_step]) np.savetxt(dir+'repeat.out',repeat_chain[:stop_step]) ### save local for debug np.savetxt(dir+'local_prob.out',local_prob_chain[:stop_step,:]) np.savetxt(dir+'local.out',local_chain[:stop_step,:]) ### save top for restart ''' top_ind = np.argsort(loglike_chain)[-100:]
hyper_domain=hyper_domain, local_domain=local_domain, \ trial_upbound = n_step*10) ''' ### inverse sampling hyper_prob0 = np.array([0.5, 0.5, 0.5]) local_prob0 = 0.5 * np.ones(2*n_group) hyper_stepsize = np.array([1e-3, 1e-3, 1e-3]) local_stepsize = 1e-3 * np.ones(2*n_group) hyper_prob_chain, hyper_chain, local_prob_chain, local_chain, \ loglikelihood_chain, repeat_chain, stop_step = \ hbm_joint_cdf(hyper_prob0, hyper_stepsize, local_prob0, local_stepsize, n_step,\ inverse_hyper, inverse_local,\ loglike_function, data=[mass_ob, mass_err, rad_ob, rad_err],\ trial_upbound = n_step*10) print 'end', time.asctime() ### plot row = 2 col = 4 f, ((a00,a01,a02,a03),(a10,a11,a12,a13))=plt.subplots(row,col,figsize=(col*5,row*5)) ax = ((a00,a01,a02,a03),(a10,a11,a12,a13)) # local chain choice = np.random.choice(np.arange(n_group),5)