unlensed = mg.get_map()
noise_map = ng.get_map()
lensed = enlensing.displace_map(unlensed, alpha_pix, order=lens_order)
tot_beamed = maps.filter_map(lensed,kbeam) #+ fg_true
stamp = tot_beamed + noise_map
if task==0:
io.plot_img(unlensed,pout_dir + "0_unlensed.png")
io.plot_img(lensed,pout_dir + "1_lensed.png")
io.plot_img(fg,pout_dir + "2_fg.png")
io.plot_img(stamp,pout_dir + "3_tot.png")
# Bayesian
totlnlikes = []
for k,kamp in enumerate(bkamps):
lnlike = maps.get_lnlike(cinvs[k],stamp) + logdets[k]
totlnlike = lnlike #+ lnprior[k]
totlnlikes.append(totlnlike)
nlnlikes = -0.5*np.array(totlnlikes)
mstats.add_to_stats("totlikes",nlnlikes)
# lnlikes2d = np.zeros((bkamps.size,famps.size))
# for k,kamp in enumerate(bkamps):
# for j,famp in enumerate(famps):
# cinv_updated, det_updated = cupdater.get_cinv(k,famp)
# lnlike = maps.get_lnlike(cinv_updated,stamp) + det_updated
# lnlikes2d[k,j] = lnlike
# mstats.add_to_stack("lnlike2d",-0.5*lnlikes2d)
enlensing.displace_map(unlensed.copy(), alpha_pix, order=lens_order),
kbeam)
fdownsampled = enmap.enmap(resample.resample_fft(lensed, bshape), bwcs)
stamp = fdownsampled + noise_map
#cutout = lensed + noise_map
cutout = stamp[int(bshape[0] / 2. - shape[0] / 2.):int(bshape[0] / 2. +
shape[0] / 2.),
int(bshape[0] / 2. - shape[0] / 2.):int(bshape[0] / 2. +
shape[0] / 2.)]
# print(cinvs[k].shape,cutout.shape)
totlnlikes = []
for k, kamp in enumerate(kamps):
lnlike = maps.get_lnlike(cinvs[k], cutout) + logdets[k]
totlnlike = lnlike #+ lnprior[k]
totlnlikes.append(totlnlike)
nlnlikes = -0.5 * np.array(totlnlikes)
mstats.add_to_stats("totlikes", nlnlikes)
mstats.get_stats()
lnlikes = mstats.vectors["totlikes"].sum(axis=0)
lnlikes -= lnlikes.max()
pl = io.Plotter(xlabel="$A$", ylabel="$\\mathrm{ln}\\mathcal{L}$")
for j in range(mstats.vectors["totlikes"].shape[0]):
pl.add(
kamps, mstats.vectors["totlikes"][j, :] /
mstats.vectors["totlikes"][j, :].max())