def exp_order(mom):
return {'calc':0, 'Linear':1, 'ZPHI':2, 'SC':3, 'MSC':4}[mom.source]
# <codecell>
for lam_text in wavelengths:
for exp in exps:
data = data_cache[lam_text, exp]
with datatypes.PlotInfoContext(wavelength=data.wavelength):
with defaults.colorbarLabeller(algLabelsNoSrc):
for pol,mom in [('H', datatypes.Attenuation), ('H', datatypes.SpecAttenuation),
(None, datatypes.DiffAtten), (None, datatypes.SpecDiffAtten)]:
fig = plt.figure(figsize=(8, 6))
moments = data.fields.grabAll(mom, filt=lambda f: f.pol==pol and f.source not in ('average', 'ts'))
moments = sorted(moments, key=exp_order)
grid = defaults.multipanel_cbar_row(fig, (1, len(moments)), moments, data,
rect=[0.07, 0.10, 0.91, 0.91])
for ax,m in zip(grid, moments):
src = m.source
l = ax.set_title('Truth' if src == 'calc' else src)
l.set_verticalalignment('bottom')
fig.suptitle('%s - %s' % (exp, mom.name), fontsize=18)
fig.savefig('%s_%s_%s.png' % (lam_text, exp, mom.name.replace(' ', '_')))
plt.close(fig)
def calc_differences_algs(data, dt=datatypes.Attenuation, pol='H'):
destMap = {datatypes.Attenuation:AttenDelta, datatypes.DiffAtten:DiffAttenDelta,
datatypes.SpecAttenuation:SpecAttenDelta, datatypes.SpecDiffAtten:SpecDiffAttenDelta}
ref_field = data.fields.grabData(dt, pol=pol, source='calc')
fields = data.fields.grabAll(dt, filt=lambda f: f.pol == pol and f.source not in ('calc', 'average', 'ts'))
for f in fields:
def make_figs(data_cache, out_dir='.'):
wavelengths,exps = data_cache.sub_keys()
process_all_atten(data_cache)
plt.rcParams['savefig.dpi'] = 200
plt.rcParams['font.size'] = 10
plots.defaults.axisDefaults.setup = setupDefaults
for lam_text in wavelengths:
for exp in exps:
data = data_cache[lam_text, exp]
with datatypes.PlotInfoContext(wavelength=data.wavelength):
with defaults.colorbarLabeller(algLabelsNoSrc):
for pol,mom in [('H', datatypes.Attenuation),
('V', datatypes.Attenuation),
(None, datatypes.DiffAtten)]:
fig = plt.figure(figsize=(6, 4))
moments = data.fields.grabAll(mom, filt=lambda f: f.pol==pol and f.source not in ('average', 'ts'))
moments = sorted(moments, key=exp_order)
grid = defaults.multipanel_cbar_row(fig, (1, len(moments)), moments, data,
rect=[0.08, 0.10, 0.91, 0.91])
for ax,m in zip(grid, moments):
src = m.source
l = ax.set_title('Truth' if src == 'calc' else src)
l.set_verticalalignment('bottom')
fig.suptitle('%s - %s' % (exp, mom.name), fontsize=14)
if pol is None:
fname = '%s_%s_%s.png' % (lam_text, exp.replace(' ', ''), mom.name.replace(' ', '_'))
else:
fname = '%s_%s_%s_%s.png' % (lam_text, exp.replace(' ', ''), mom.name.replace(' ', '_'), pol)
fig.savefig(os.path.join(out_dir, fname), bbox_inches='tight')
plt.close(fig)
def calc_differences_algs(data, dt=datatypes.Attenuation, pol='H'):
destMap = {datatypes.Attenuation:datatypes.AttenDelta,
datatypes.DiffAtten:datatypes.DiffAttenDelta,
datatypes.SpecAttenuation:datatypes.SpecAttenDelta,
datatypes.SpecDiffAtten:datatypes.SpecDiffAttenDelta}
ref_field = data.fields.grabData(dt, pol=pol, source='calc')
fields = data.fields.grabAll(dt, filt=lambda f: f.pol == pol and f.source not in ('calc', 'average', 'ts'))
for f in fields:
data.addField(data.fields[f] - ref_field, destMap[dt], pol=f.pol, source=f.source)
for d in data_cache.values():
calc_differences_algs(d, dt=datatypes.Attenuation, pol='H')
calc_differences_algs(d, dt=datatypes.Attenuation, pol='V')
#calc_differences_algs(d, dt=datatypes.SpecAttenuation, pol='H')
calc_differences_algs(d, dt=datatypes.DiffAtten, pol=None)
#calc_differences_algs(d, dt=datatypes.SpecDiffAtten, pol=None)
for lam_text in wavelengths:
for exp in exps:
data = data_cache[lam_text, exp]
with datatypes.PlotInfoContext(wavelength=data.wavelength):
with defaults.colorbarLabeller(algLabelsNoSrc):
for pol,mom in [('H', datatypes.AttenDelta),
('V', datatypes.AttenDelta),
(None, datatypes.DiffAttenDelta)]:
fig = plt.figure(figsize=(6, 4))
moments = data.fields.grabAll(mom, filt=lambda f: f.pol==pol and f.source not in ('average', 'ts'))
moments = sorted(moments, key=exp_order)
grid = defaults.multipanel_cbar_row(fig, (1, len(moments)), moments, data,
rect=[0.08, 0.05, 0.91, 0.85])
for ax,m in zip(grid, moments):
l = ax.set_title(m.source)
l.set_verticalalignment('bottom')
fig.suptitle('%s - %s' % (exp, mom.name), fontsize=14)
if pol is None:
fname = '%s_%s_%s.png' % (lam_text, exp.replace(' ', ''), mom.name.replace(' ', '_'))
else:
fname = '%s_%s_%s_%s.png' % (lam_text, exp.replace(' ', ''), mom.name.replace(' ', '_'), pol)
fig.savefig(os.path.join(out_dir, fname), bbox_inches='tight')
plt.close(fig)
limits = {('C', 'H'):2.0, ('C', 'V'):2.0, ('C', None):0.75,
('X', 'H'):8.0, ('X', 'V'):8.0, ('X', None):1.0}
#stats = StatsFile(os.path.join(out_dir, 'spec_atten_stats.tex'))
for lam_text in wavelengths:
for exp in exps:
#with stats.table(lam_text, exp) as tab:
data = data_cache[lam_text, exp]
with datatypes.PlotInfoContext(wavelength=data.wavelength):
with defaults.colorbarLabeller(algLabelsNoSrc):
for pol,mom in [('H', datatypes.SpecAttenuation),
('V', datatypes.SpecAttenuation),
(None, datatypes.SpecDiffAtten)]:
ref_mom = data.fields.grab(mom, pol=pol, source='calc')
ref_field = data.fields[ref_mom]
moments = data.fields.grabAll(mom,
filt=lambda f: f.pol == pol and f.source not in ('calc', 'average', 'ts', 'sweep'))
moments = sorted(moments, key=exp_order)
nrows = 2
ncols = len(moments) // nrows
fig, axes = plt.subplots(nrows, ncols, sharex=True,
sharey=True, figsize=(6, 4))
fig.subplots_adjust(left=0.10, top=0.85, bottom=0.12,
hspace=0.45)
xlim = limits[lam_text, pol]
for ind, (m, ax, panel_label) in enumerate(zip(moments,
axes.flatten(), LabelGenerator('a'))):
alg_field = data.fields[m]
mom_abbr = m.string_parts()[0]
mask = (ref_field > 0.02) & (alg_field > 0.02)
norm = datatypes.TypePlotInfo[ref_mom.type]['norm']
hist, xedge, yedge = np.histogram2d(ref_field[mask], alg_field[mask], bins=50,
range=[[norm.vmin, norm.vmax], [norm.vmin, norm.vmax]])
ax.pcolormesh(xedge, yedge, hist.T, norm=LogNorm())
ax.plot([0.0, xlim], [0.0, xlim], 'k--')
ax.grid(True)
panel_label.patch.set_boxstyle("round, pad=0., rounding_size=0.2")
ax.add_artist(panel_label)
l = ax.set_title(m.source)
l.set_verticalalignment('bottom')
if ind % nrows == 0:
ax.set_ylabel(mom_abbr)
ax.set_xlabel('True ' + mom_abbr)
stats[lam_text, exp, pol, m.source] = regress_stats(ref_field[mask], alg_field[mask])
axes[0,0].set_xlim(0.0, xlim)
axes[0,0].set_ylim(0.0, xlim)
axes[0,0].xaxis.set_major_locator(plt.MaxNLocator(nbins=5))
axes[0,0].yaxis.set_major_locator(plt.MaxNLocator(nbins=5))
fig.suptitle('%s - %s' % (exp, mom.name), fontsize=14)
if pol is None:
fname = '%s_%s_%s_scatter.png' % (lam_text, exp.replace(' ', ''), mom.name.replace(' ', '_'))
else:
fname = '%s_%s_%s_%s_scatter.png' % (lam_text, exp.replace(' ', ''), mom.name.replace(' ', '_'), pol)
fig.savefig(os.path.join(out_dir, fname), bbox_inches='tight')
plt.close(fig)
names = ['Bias', 'MSE', '$r^2$']
with open('spec_atten_stats.tex', 'w') as fp:
for lam_text in wavelengths:
for pol,abbr in [('Horizontal','H'), ('Vertical', 'V'), ('Differential', None)]:
for ind,name in enumerate(names):
fp.write('''\\begin{{frame}}
\\frametitle{{Summary: {name} at {lam}-band for {pol} Polarization}}
\\begin{{center}}
\\begin{{tabular}}{{| c | c | c | c | c |}}
\\hline
Experiment & Linear & ZPHI & SC & MSC \\\\
\\hline
\\hline\n'''.format(name=name, lam=lam_text, pol=pol))
for exp in exps:
fp.write(' %s & %.4f & %.4f & %.4f & %.4f \\\\\n'%(exp,
stats[lam_text, exp, abbr, 'Linear'][ind],
stats[lam_text, exp, abbr, 'ZPHI'][ind],
stats[lam_text, exp, abbr, 'SC'][ind],
stats[lam_text, exp, abbr, 'MSC'][ind]))
fp.write(r''' \hline
\end{tabular}
\end{center}
\end{frame}
''')