# HACKY: Temporarily set evaluate.USEBINS to be all true to use *all* angular bins for
# calculating the product moment correlation coeff
evaluate.USEBINS = np.ones(evaluate.USEBINS.shape, dtype=bool)
# Get truth as a full catalog
_, x, y, g1true, g2true = test_evaluate.get_variable_gtrue(
EXPERIMENT, OBS_TYPE, truth_dir=TRUTH_DIR)
# Then make a variable submission to learn what pure c1 and c2 look like in map^2
import tempfile
fdtmp, tmpfile = tempfile.mkstemp(suffix=".dat")
result = test_evaluate.make_variable_submission(x,
y,
np.zeros_like(g1true),
np.zeros_like(g2true),
np.zeros_like(g1true),
np.zeros_like(g2true),
1.,
0.,
0.,
0.,
outfile=tmpfile,
noise_sigma=0.)
os.close(fdtmp)
map_E_c1 = np.loadtxt(tmpfile)[:, 2]
os.remove(tmpfile)
fdtmp, tmpfile = tempfile.mkstemp(suffix=".dat")
result = test_evaluate.make_variable_submission(x,
y,
np.zeros_like(g1true),
np.zeros_like(g2true),
np.zeros_like(g1true),
np.zeros_like(g2true),
EXPERIMENT, obs_type, truth_dir=TRUTH_DIR)
_, _, _, g1int, g2int = test_evaluate.get_variable_gsuffix(
EXPERIMENT, obs_type, truth_dir=TRUTH_DIR)
for jc, cval in enumerate(CVALS):
for itest in xrange(NTEST):
# Build the submission
fdsub, subfile = tempfile.mkstemp(suffix=".dat")
result = test_evaluate.make_variable_submission(
x,
y,
g1true,
g2true,
g1int,
g2int,
cval,
cval,
evaluate.MFID,
evaluate.MFID,
outfile=subfile,
noise_sigma=NOISE_SIGMA[obs_type])
os.close(fdsub)
# Then evaluate Q_v
qc[obs_type][itest, jc] = evaluate.q_variable(
subfile,
EXPERIMENT,
obs_type,
usebins=evaluate.USEBINS,
truth_dir=TRUTH_DIR)
os.remove(subfile)
_, x, y, g1true, g2true = test_evaluate.get_variable_gtrue(
EXPERIMENT, obs_type, truth_dir=TRUTH_DIR)
_, _, _, g1int, g2int = test_evaluate.get_variable_gsuffix(
EXPERIMENT, obs_type, truth_dir=TRUTH_DIR)
# Then loop over specified number of tests
for itest in xrange(NTEST):
# Build the submission
fdsub, subfile = tempfile.mkstemp(suffix=".dat")
result = test_evaluate.make_variable_submission(
x,
y,
g1true,
g2true,
g1int,
g2int,
0.,
0.,
0.,
0.,
outfile=subfile,
noise_sigma=NOISE_SIGMA[obs_type])
os.close(fdsub)
data = np.loadtxt(subfile)
mapE[obs_type][:, itest] = data[:, 2]
print "Completed test %4d / %4d" % (itest + 1, NTEST)
os.remove(subfile)
print "Saving pickled map_E tables to " + mapEoutfile
with open(mapEoutfile, "wb") as fout:
cPickle.dump(mapE, fout)
print "NOISE_SIGMA = "+str(NOISE_SIGMA[obs_type])
# First we build the truth table
print "Building truth tables for control-"+obs_type+"-variable"
# Get the x,y, true intrinsic ellips and shears for making fake submissions
_, x, y, g1true, g2true = test_evaluate.get_variable_gtrue(
EXPERIMENT, obs_type, truth_dir=TRUTH_DIR)
_, _, _, g1int, g2int = test_evaluate.get_variable_gsuffix(
EXPERIMENT, obs_type, truth_dir=TRUTH_DIR)
for jc, cval in enumerate(CVALS):
for itest in xrange(NTEST):
# Build the submission
fdsub, subfile = tempfile.mkstemp(suffix=".dat")
result = test_evaluate.make_variable_submission(
x, y, g1true, g2true, g1int, g2int, cval, cval, evaluate.MFID,
evaluate.MFID, outfile=subfile, noise_sigma=NOISE_SIGMA[obs_type])
os.close(fdsub)
# Then evaluate Q_v
qc[obs_type][itest, jc] = evaluate.q_variable(
subfile, EXPERIMENT, obs_type, usebins=evaluate.USEBINS,
truth_dir=TRUTH_DIR)
os.remove(subfile)
print "Test %4d / %4d (c = %.3e) Q_v = %.4f" % (
itest+1, NTEST, cval, qc[obs_type][itest, jc])
print "Mean Q_v = "+str(qc[obs_type][:, jc].mean())+" for "+str(NTEST)+\
" sims (with c = "+str(cval)+", obs_type = "+str(obs_type)+")"
print
print "Saving pickled Q_v versus c dict to "+coutfile
print "Calculating map_E values for control-"+obs_type+"-variable data in GREAT3"
print "NOISE_SIGMA = "+str(NOISE_SIGMA[obs_type])
# First we build the truth table
print "Building truth tables for control-"+obs_type+"-variable"
# Get the x,y, true intrinsic ellips and shears for making fake submissions
_, x, y, g1true, g2true = test_evaluate.get_variable_gtrue(
EXPERIMENT, obs_type, truth_dir=TRUTH_DIR)
_, _, _, g1int, g2int = test_evaluate.get_variable_gsuffix(
EXPERIMENT, obs_type, truth_dir=TRUTH_DIR)
# Then loop over specified number of tests
for itest in xrange(NTEST):
# Build the submission
fdsub, subfile = tempfile.mkstemp(suffix=".dat")
result = test_evaluate.make_variable_submission(
x, y, g1true, g2true, g1int, g2int, 0., 0., 0., 0., outfile=subfile,
noise_sigma=NOISE_SIGMA[obs_type])
os.close(fdsub)
data = np.loadtxt(subfile)
mapE[obs_type][:, itest] = data[:, 2]
print "Completed test %4d / %4d" % (itest+1, NTEST)
os.remove(subfile)
print "Saving pickled map_E tables to "+mapEoutfile
with open(mapEoutfile, "wb") as fout:
cPickle.dump(mapE, fout)
print
else:
with open(mapEoutfile, "rb") as fin:
mapE = cPickle.load(fin)
# Hmmm actually a linear model doesn't work because we do want to constrain c**2 to be
# positive... Try using optimize instead
if USE_ALL_BINS:
# HACKY: Temporarily set evaluate.USEBINS to be all true to use *all* angular bins for
# calculating the product moment correlation coeff
evaluate.USEBINS = np.ones(evaluate.USEBINS.shape, dtype=bool)
# Get truth as a full catalog
_, x, y, g1true, g2true = test_evaluate.get_variable_gtrue(
EXPERIMENT, OBS_TYPE, truth_dir=TRUTH_DIR)
# Then make a variable submission to learn what pure c1 and c2 look like in map^2
import tempfile
fdtmp, tmpfile = tempfile.mkstemp(suffix=".dat")
result = test_evaluate.make_variable_submission(
x, y, np.zeros_like(g1true), np.zeros_like(g2true), np.zeros_like(g1true),
np.zeros_like(g2true), 1., 0., 0., 0., outfile=tmpfile, noise_sigma=0.)
os.close(fdtmp)
map_E_c1 = np.loadtxt(tmpfile)[:, 2]
os.remove(tmpfile)
fdtmp, tmpfile = tempfile.mkstemp(suffix=".dat")
result = test_evaluate.make_variable_submission(
x, y, np.zeros_like(g1true), np.zeros_like(g2true), np.zeros_like(g1true),
np.zeros_like(g2true), 0., 1., 0., 0., outfile=tmpfile, noise_sigma=0.)
os.close(fdtmp)
map_E_c2 = np.loadtxt(tmpfile)[:, 2]
os.remove(tmpfile)
# Then average these (I checked they are very similar as you would expect) to get our "unit c"
# term for the modelling. Note that previously I summed these, but I think this was an error as
# that actually corresponds approximately to a shear field with g1 = g2 = 1, i.e. |g| = sqrt(2)
map_E_unitc = .5 * (map_E_c1 + map_E_c2)
